advanced science letters - istieistie.org/21asl10-pages-proofsneedcorrections.pdf · james r....

SUBMISSION OF MANUSCRIPT

Authors should submit manuscript electronically either as a PDF or Microsoft Word file to the Editor-in-Chief, or one of the Regional Editors, or one of the Associate Editors listed above. Authors are highly encouraged to submit manuscript electronically to save time for the reviewing process.

ADVANCED SCIENCE LETTERSVOLUME 21, NUMBER 10 2947–3429 (2015)

OCTOBER 2015 ISSN: 1936-6612, EISSN: 1936-7317

www.aspbs.com/science

EDITORIAL BOARD

Filippo Aureli, UK

Marcel Ausloos, Belgium

Martin Bojowald, USA

Sougato Bose, UK

Jacopo Buongiorno, USA

Paul Cordopatis, Greece

Maria Luisa Dalla Chiara, Italy

Dionysios Demetriou Dionysiou, USA

Simon Eidelman, Russia

Norbert Frischauf, Austria

Toshi Futamase, Japan

Leonid Gavrilov, USA

Vincent G. Harris, USA

Mae-Wan Ho, UK

Keith Hutchison, Australia

David Jishiashvili, Georgia

George Khushf, USA

Sergei Kulik, Russia

Harald Kunstmann, Germany

Alexander Lebedev, Russia

James Lindesay, USA

Michael Lipkind, Israel

Nigel Mason, UK

Johnjoe McFadden, UK

B. S. Murty, India

Heiko Paeth, Germany

Matteo Paris, Italia

David Posoda, Spain

Paddy H. Regan, UK

Leonidas Resvanis, Greece

Wolfgang Rhode, Germany

Derek C. Richardson, USA

Carlos Romero, Brazil

Andrea Sella, UK

P. Shankar, India

Surya Singh, UK

Leonidas Sotiropoulos, Greece

Roger Strand, Norway

Karl Svozil, Austria

Kit Tan, Denmark

Roland Triay, France

Rami Vainio, Finland

Victor Voronov, Russia

Andrew Whitaker, Ireland

Lijian Xu, China

Alexander Yefremov, Russia

Avraam Zelilidis, Greece

Alexander V. Zolotaryuk, Ukraine

HONORARY EDITORS

Richard Ernst*, ETH Zürich, Switzerland

Eric B. Karlsson**, Uppsala University, Sweden

Douglas Osheroff*, Stanford University, USA*Nobel Prize Laureate **Member of the Nobel Committee for Physics 1987–1998 (chairman in 1998)

EUROPEAN EDITOR

Prof. Dr. Wolfram SchommersForschungszentrum Karlsruhe, Institut für Wissenschaftliches Rechnen, D-76021 Karlsruhe, GERMANYTel.: +49-7247-82-2432; Fax: +49-7247-82-4972; E-mail: [email protected]

ASIAN EDITOR

Dr. Katsuhiko Ariga, Ph.D.Advanced Materials Laboratory, National Institute for Materials Science, 1-1 Namiki, Tsukuba, Ibaraki 305-0044, JAPAN

Tel.: +81-29-860-4832; Fax: +81-29-860-4832; E-mail: [email protected]

Editor-in-Chief: Dr. Hari Singh Nalwa, Ph.D.American Scientific Publishers, 26650 The Old Road, Suite 208, Valencia, California 91381-0751, USA

Phone: (661) 799-7200 Fax: (661) 799-7230 E-mail: [email protected]

Diederik Aerts, Belgium

Yakir Aharonov, Israel

Peter C. Aichelburg, Austria

Jim Al-Khalili, UK

Simon Baron-Cohen, UK

Jake Blanchard, USA

Franz X. Bogner, Germany

John Borneman, USA

John Casti, Austria

Masud Chaichian, Finland

Sergey V. Chervon, Russia

Kevin Davey, USA

Tania Dey, Canada

Frans de Waal, USA

Roland Eils, Germany

Marco Genovese, Italia

Bert Gordijn, The Netherlands

Thomas Görnitz, Germany

Ji-Huan He, China

Nongyue He, China

Irving P. Herman, USA

Dipankar Home, India

Jucundus Jacobeit, Germany

Yuriy A. Knirel, Russia

Arthur Konnerth, Germany

G. A. Kourouklis, Greece

Peter Krammer, Germany

Andrew F. Laine, USA

Minbo Lan, China

Martha Lux-Steiner, Germany

Klaus Mainzer, Germany

JoAnn E. Manson, USA

Mark P. Mattson, USA

Lucio Mayer, Switzerland

Efstathios Meletis, USA

Karl Menten, Germany

Yoshiko Miura, Japan

Fred M. Mueller, USA

Garth Nicolson, USA

Nina Papavasiliou, USA

Panos Photinos, USA

Constantin Politis, Greece

Zhiyong Qian, China

Reinhard Schlickeiser, Germany

Surinder Singh, USA

Suprakas Sinha Ray, South Africa

Koen Steemers, UK

Shinsuke Tanabe, Japan

James R. Thompson, USA

Uwe Ulbrich, Germany

Ahmad Umar, Saudi Arabia

ASSOCIATE EDITORS

00_21ASL10-FM.indd 200_21ASL10-FM.indd 2 12/1/2015 3:26:45 PM12/1/2015 3:26:45 PM

ADVANCED SCIENCE LETTERSVOLUME 21, NUMBER 10 2947–3429 (2015)

OCTOBER 2015 ISSN: 1936-6612, EISSN: 1936-7317

www.aspbs.com/science

A SPECIAL ISSUE

2947–2951 Selected Peer-Reviewed Articles from the 3rd International Conference on Internet Services Technology and Information Engineering 2015 (ISTIE 2015)

Guest Editors: Ford Lumban Gaol and Benfano Soewito

Adv. Sci. Lett. 21, 2947–2951 (2015)

REVIEWS

2952–2956 A Review on Feature Selection Methods for Sentiment Analysis

Lai Po Hung, Rayner Alfred, and Mohd Hanafi Ahmad Hijazi

Adv. Sci. Lett. 21, 2952–2956 (2015)


2957–2962 A Review on the Ensemble Framework for Sentiment Analysis

Lai Po Hung, Rayner Alfred, Mohd Hanafi Ahmad Hijazi, and Ag. Asri Ag. Ibrahim

Adv. Sci. Lett. 21, 2957–2962 (2015)

2963–2966 ICT Intervention and Palliative Care: A Review

Noor Azizah Mohamadali, Adebiyi Lookman Ademola, Mira Kartiwi, and Zainatul Shima

Adv. Sci. Lett. 21, 2963–2966 (2015)

RESEARCH ARTICLES

2967–2973 Low Loss Magnetic Design for Wireless Power Transfer

Heinz Zenkner and Werachet Khan-Ngern

Adv. Sci. Lett. 21, 2967–2973 (2015)

2974–2978 Design of a Quality of Experience-Based Digital Convergence iHome

Hao-Hsiang Ku and Ching-Ho Chi

Adv. Sci. Lett. 21, 2974–2978 (2015)

2979–2984 The Study of Wireless Power Transmission Under Water-to-Air Mode

Werachet Khan-Ngern and Heinz Zenkner

Adv. Sci. Lett. 21, 2979–2984 (2015)

2985–2988 Study on Optimization of Valve Parameter for Multi-Cylinder Synchronous Control System

Gwang Seok Kim and Deok Jin Lee

Adv. Sci. Lett. 21, 2985–2988 (2015)

Ferrite

Strands, twisted


2989–2992 A Wireless Water Flow Monitoring in a Closed Channel Pipeline for Leakage Detection

Anif Jamaluddin, Pety Refiyanti, Dewanto Harjunowibowo, Lita Rahmasari, Jamzuri, R. Dwi Teguh, and A. Priyo Heru

Adv. Sci. Lett. 21, 2989–2992 (2015)

2993–2996 Precise Indoor Positioning Algorithms for Autonomous Hexa-Rotor Using Cubature Kalman Filter

Byeongju Kang, KilTo Chong, and Deokjin Lee

Adv. Sci. Lett. 21, 2993–2996 (2015)

2997–3001 Computational Modeling of Mood from Sequence of Emotions

Dini Handayani, Hamwira Yaacob, Abdul Wahab Abdul Rahman, Wahju Sediono, and Asadullah Shah

Adv. Sci. Lett. 21, 2997–3001 (2015)

3002–3006 Making a Successful Agile Team

Beni Suranto

Adv. Sci. Lett. 21, 3002–3006 (2015)

3007–3009 Optimized Performance Result for 2.4 GHz and 2.45 GHz Circularly Polarized Microstrip Antenna

Rudy Yuwono, Ronanobelta Syakura, Erni Yudaningtyas, Endah B. Purnomowati, and Aisah

Adv. Sci. Lett. 21, 3007–3009 (2015)

0 1 2 3 4 5

16,5

17,0

17,5

18,0

18,5

19,0

Rat

e F

low

(L/

Min

)

Number of Hole (Leakage Source)

The Graph Number of Hole Vs Rate Flow


3010–3014 Cultural-Influenced Speech Emotion Recognition System Using Multi Layer Perceptron and Support Vector Machine

Norhaslinda Kamaruddin, Abdul Wahab, and Anis Abd. Kamal Sayuti

Adv. Sci. Lett. 21, 3010–3014 (2015)

3015–3019 Indoor 3D Mapping Technique Based on 3D Image Transformation Using Kinect Sensor

Doopalam Tuvshinjargal, Byeongju Kang, Kil To Chong, and Deok Jin Lee

Adv. Sci. Lett. 21, 3015–3019 (2015)

3020–3024 A New Technique for Protecting Server Against MAC Spoofing via Software Attestation

Kamarularifin Abd Jalil, Nor Shahniza Kamal Bashah, and Mohd Hariz Naim @ Mohayat

Adv. Sci. Lett. 21, 3020–3024 (2015)

3025–3029 Precursor Emotion of Driver by Using Electroencephalogram (EEG) Signals

Norzaliza Md Nor and Abdul Wahab Bar

Adv. Sci. Lett. 21, 3025–3029 (2015)

3030–3033 Optimized Walficsh-Bertoni Model for Path Loss Prediction DTTV Propagation in Urban Area of Southern Thailand

Pitak Keawbunsong, Pitchaya Supannakoon, and Sathaporn Promwong

Adv. Sci. Lett. 21, 3030–3033 (2015)

3034–3037 Mobile Navigation System Using Fuzzy C-Mean Clustering and Subtractive Clustering Based on Fingerprinting Technique

Jirapat Sangthong and Sathaporn Promwong

Adv. Sci. Lett. 21, 3034–3037 (2015)

FearSad

12

3

4ht

HB

dR


3038–3042 A Performance Comparison of Statistical and Machine Learning Techniques in Learning Time Series Data

Haviluddin, Rayner Alfred, Joe Henry Obit, Mohd Hanafi Ahmad Hijazi, and Ag Asri Ag Ibrahim

Adv. Sci. Lett. 21, 3038–3042 (2015)

3043–3046 Effect of Vortex Order on Helical-Phased Donut Mode Launch in Multimode Fiber

Angela Amphawan, Yousef Fazea, Tarek Elfouly, and Khalid Abualsaud

Adv. Sci. Lett. 21, 3043–3046 (2015)

3047–3050 Optical Mode Division Multiplexing for Secure Ro-FSO WLANs

Angela Amphawan, Sushank Chaudhary, Tarek Elfouly, and Khalid Abualsaud

Adv. Sci. Lett. 21, 3047–3050 (2015)

3051–3054 Hermite-Gaussian Mode Division Multiplexing for Free-Space Optical Interconnects

Angela Amphawan, Sushank Chaudhary, and Tse-Kian Neo

Adv. Sci. Lett. 21, 3051–3054 (2015)

3055–3059 Radio Subcarrier Spacing Effect on SCM-MDM Using HG Modes in Radio-Over-Fiber

Baseem Khalaf Alsharaa, Angela Amphawan, and Tse-Kian Neo

Adv. Sci. Lett. 21, 3055–3059 (2015)


3060–3064 Analysis of Data Quality Maturity in a Higher Education Institution

W. Yohana Dewi Lulu, Adhistya Erna Permanasari, Ridi Ferdiana, and Lukito Edi Nugroho

Adv. Sci. Lett. 21, 3060–3064 (2015)

3065–3069 Optimization of Path Loss Model for Prediction DTTV Propagation in Urban Area of Southern Thailand

Pitak Keawbunsong, Pitchaya Supannakoon, and Sathaporn Promwong

Adv. Sci. Lett. 21, 3065–3069 (2015)

3070–3074 Context-Aware Mobile Learning Model for Traveler

Dadang Syarif Sihabudin Sahid, Lukito Edi Nugroho, Ridi Ferdiana, and Paulus Insap Santosa

Adv. Sci. Lett. 21, 3070–3074 (2015)

3075–3079 Oriented Bounding Box Optimization on the Rotation Group SO(3, ) Based on Particle Swarm Optimization-Nelder Mead

Taslimatul Atsna Faizati, Guruh Fajar Shidik, and Vincent Suhartono

Adv. Sci. Lett. 21, 3075–3079 (2015)

3080–3084 Prediction of CO2 Emissions Using an Artificial Neural Network: The Case of the Sugar Industry

Chairul Saleh, Raden Achmad Chairdino Leuveano, Mohd Nizam Ab Rahman, Baba Md Deros, and Nur Rachman Dzakiyullah

Adv. Sci. Lett. 21, 3080–3084 (2015)

2 2.5 3 3.5 4 4.5 5 5.5 6 6.5 7110

115

120

125

130

135

140

145

150

155

160

Distance (Km)

Pat

h lo

ss (

dB)

Linear regression fit : AverageOptimized Hata : CH26Optimized Hata : CH42Optimized Hata : CH46Optimized Hata : CH54Old Hata Model

y = 3.26*x + 117.79

0.000

0.200

0.400

0.600

0.800

1.000

1.200

1 9 17 25 33 41 49 57 65 73 81 89 97 105

113

121

Err

or V

alue

Data

Actual

Prediction


3085–3088 Classification of Resting State Electroencephalography for the Identification of Asperger’s Syndrome

Nurul Izzati Mat Razi, Marini Othman, and Abdul Wahab

Adv. Sci. Lett. 21, 3085–3088 (2015)

3089–3092 Identification of Pure and Impure Palmyrah Palm Juice by Using Microwave Sensor System

Rachen Kanahna, Panisa Keowsawat, Sathaporn Promwong, and Chuwong Phongcharoenpanich

Adv. Sci. Lett. 21, 3089–3092 (2015)

3093–3099 Semantic Graph Application to Call Center for Entity-Relation Search

Takahiro Kawamura and Akihiko Ohsuga

Adv. Sci. Lett. 21, 3093–3099 (2015)

3100–3102 A Resilient Exponential Smoothing Method for Link Speed Forecasting

Kyung-Il Choe

Adv. Sci. Lett. 21, 3100–3102 (2015)

3103–3106 Semantic Representation of Virtual Humans

Mezati Messaoud, Foudil Cherif, Cédric Sanza, and Véronique Gaildrat

Adv. Sci. Lett. 21, 3103–3106 (2015)

3107–3110 Integrating Cognitive Load Theory in Video Based Learning Environment

Nur Khairiyah Kadar, Norah Md Noor, and Juhazren Junaidi

Adv. Sci. Lett. 21, 3107–3110 (2015)

–30

–35

–40

–45

–50

–5518 20 22 24 26 28 30 32

Total Soluble Solid (°Brix)

|S21

| (dB

) Microwave sensor

Network analyzer

pure juice

with granulated sugarwith coconut sugar

MAE

MAPE

MAE MAPE


3111–3114 A Business Intelligence-Driven Approach to Government Enterprise Architecture

Teduh Dirgahayu

Adv. Sci. Lett. 21, 3111–3114 (2015)

3115–3118 FWAT (Fast Wartegg Analyzer Tool) for Personality Identification

Rosihan Ari Yuana, Dewanto Harjunowibowo, and Nugroho Karyanta

Adv. Sci. Lett. 21, 3115–3118 (2015)

3119–3123 Factors Persuading Nuts and Bolts of Agile Estimation

Saru Dhir and Deepak Kumar

Adv. Sci. Lett. 21, 3119–3123 (2015)

3124–3128 Graph Extraction Algorithm for Volumetric Segmentation

Dumitru Dan Burdescu, Liana Stanescu, Marius Brezovan, Cosmin Stoica Spahiu, and Daniel Costin Ebanca

Adv. Sci. Lett. 21, 3124–3128 (2015)

3129–3132 Predicting Traffic Accident Severity Using Classification Techniques

Seok-Lyong Lee

Adv. Sci. Lett. 21, 3129–3132 (2015)

3133–3137 Modeling Product Design Knowledge

Haryani Haron, Mohd Nazri Mustafa, M. Hamiz, and Nor Diana Ahmad

Adv. Sci. Lett. 21, 3133–3137 (2015)

Information architecture

Business architecture

Application architecture

Technology architecture

Architecture vision(refers to BI objectives)

1

2

3

4

Module Not ActiveFunctionality WiMAX

WiFi

Low Throughput

Connection Error


3138–3142 Identification E-Learning Readiness in the Faculty of Agricultural Technology Jambi University

Kurniabudi, SetiawanAssegaff, and Sharipuddin

Adv. Sci. Lett. 21, 3138–3142 (2015)

3143–3146 Understanding the Correlation of Explicit and Implicit Memory Effectiveness Using Electroencephalograph-Based Emotional Arousal

Khamis Faraj Alarabi, Abdul Wahab, Mariam Adawiah Dzulkifli, and Norhaslinda Kamaruddin

Adv. Sci. Lett. 21, 3143–3146 (2015)

3147–3151 An Investigation on the Use of Mobile Devices Among Older People

Sofianiza Abd Malik, Muna Azuddin, Lili Marziana Abdullah, and Murni Mahmud

Adv. Sci. Lett. 21, 3147–3151 (2015)

3152–3156 Linear Vector Quantization Algorithm for Pattern Recognition on Paper Currency’s Feature Using UV Light

Dewanto Harjunowibowo, Anif Jamaluddin, Sri Hartati, Rosihan Ari Yuana, Aris Budianto, and Farid Ahmadi

Adv. Sci. Lett. 21, 3152–3156 (2015)

3157–3161 Cloud Storage Security Based on Group Key

Parinya Natho and Pramote Kuacharoen

Adv. Sci. Lett. 21, 3157–3161 (2015)

LVQClassifyT/F

NN Process Extraction114x90 px

Result

Pre image DataAcquisition

Normalization550x240 px

GroupKey

Member1Member2

Member3

--------

-----------

---------

File 1 File 1 File n

...

Directory 1

File 1 File 1 Directory 1

Directory 2

Root directoryMasterKey

--------

-----------

---------

File n

...

GroupInfo

Share Key

KeyServer


3162–3165 The Proposed Public Key Infrastructure Authentication Framework (PKIAF) for Malaysian Government Agencies

Noraida Aman Nor, Ganthan Narayana Samy, Rabiah Ahmad, Roslina Ibrahim, and Nurazean Maarop

Adv. Sci. Lett. 21, 3162–3165 (2015)

3166–3170 Tablet Technology and Apps to Enhance Slow Learners Motivation in Learning

Azizzeanna Hassan and Murni Mahmud

Adv. Sci. Lett. 21, 3166–3170 (2015)

3171–3175 Enhanced Classification Performances of Travel and Tourism Competitiveness Model with Feature Selection

Anongnart Srivihok and Arunee Intrapairot

Adv. Sci. Lett. 21, 3171–3175 (2015)

3176–3180 Design Optimization Shape Web Opening of Cellular Steel Beams

Suharjanto

Adv. Sci. Lett. 21, 3176–3180 (2015)

3181–3185 Teaching Duet in Social Sciences Education in Promoting Critical Thinking Abilities

Nurul‘Izzati Hamizan, Norasykin Mohd Zaid, and Norah Md. Noor

Adv. Sci. Lett. 21, 3181–3185 (2015)

3186–3189 Pathloss Calculation and Analysis Using Different Carrier Frequency on Wideband Code Division Multiple Access Technology

Ir. Endah Budi Purnomowati, Gaguk Asmungi, Anindito Yusuf Wirawan, and Rudy Yuwono

Adv. Sci. Lett. 21, 3186–3189 (2015)


3190–3194 Internet Banking Transaction Authentication Using Mobile One-Time Password and QR Code

Puchong Subpratatsavee and Pramote Kuacharoen

Adv. Sci. Lett. 21, 3190–3194 (2015)

3195–3198 Characteristic of Ultra Wideband Body Area Network Channel

Sanit Teawchim and Sathaporn Promwong

Adv. Sci. Lett. 21, 3195–3198 (2015)

3199–3201 Knowledge Management Implementation of Virtual Corporate Memory in Institutions of Higher Learning

Suzana Basaruddin, Haryani Haron, Wan Nor Hannani Wan Dagang, Siti Arpah Noordin, and Nor Diana Ahmad

Adv. Sci. Lett. 21, 3199–3201 (2015)

3202–3205 A System to Identify Technological Chances in Technology-Based Services

Chulhyun Kim and Moon-Soo Kim

Adv. Sci. Lett. 21, 3202–3205 (2015)

3206–3210 Statistical Analysis Carbon Footprint in Supply Chain Management

Akhmad Fauzy, Chairul Saleh, Nashrullah Setiawan, and Luqman Hakim

Adv. Sci. Lett. 21, 3206–3210 (2015)

The transaction verification result

2 2.5 3 3.5 4 4.5 5 5.5 60

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Frequency (GHz)

Mag

nit

ud

e (V

/MH

z)

Rectangularpassband waveform

Root raised cosinepassband waveform

OperationalDatabase

(Regular case)

VCM repository(Special case)

Operationalprocess

Operational System

User

CO2 Emission CO2 Emission Prediction


3211–3214 Application of Simulation Approach to Predict Carbon Emissions in Sugar Industry

Chairul Saleh, Nashrullah Setiawan, Nur Feriyanto, Akhmad Fauzy, and Riza Nurhidayati

Adv. Sci. Lett. 21, 3211–3214 (2015)

3215–3219 Hybrid Supply Chain Operations Reference-System Dynamics Performance Measurement MTS-MTO Production Typology for Batik Industry

Taufiq Immawan, Marimin, Yandra Arkeman, and Agus Maulana

Adv. Sci. Lett. 21, 3215–3219 (2015)

3220–3225 Techniques for Improving Website Rankings with Search Engine Optimization (SEO)

Kittisak Chotikitpat, Prachyanun Nilsook, and Sunantha Sodsee

Adv. Sci. Lett. 21, 3220–3225 (2015)

3226–3230 Development of Student Data Mart Using Normalized Data Store Architecture

Luthfia Rahman, Slamet Riyadi, and Eko Prasetyo

Adv. Sci. Lett. 21, 3226–3230 (2015)

3231–3234 Security Issues of Ad-Hoc Network of Sensor Node in Disaster Mitigation Plan

Sera Syarmila Sameon and Norziana Jamil

Adv. Sci. Lett. 21, 3231–3234 (2015)

3235–3239 A New Model of StegaSVM-Shifted LSB in Discrete Cosine Transform Domain on Image Steganography Approach

Hanizan Shaker Hussain, H. Yaacob, M. Sabri, A. Azmer, and Roshidi Din

Adv. Sci. Lett. 21, 3235–3239 (2015)

Node 5

Ad HocNetworkTopologyNode 1

Node 3

Node 0

Node 2Node 4

WirelessTransceiver

Sensingand

ProcessingUnit


3240–3243 Continuous Review Probabilistic Inventory Analysis Using Type-2 Fuzzy Logic

Muhammad Ridwan Andi Purnomo

Adv. Sci. Lett. 21, 3240–3243 (2015)

3244–3248 Identifying the Level of User Awareness and Factors on Phishing Attempt Among Students

Vanisri Nagalingam, Ganthan Narayana Samy, Rabiah Ahmad, Nurazean Maarop, and Roslina Ibrahim

Adv. Sci. Lett. 21, 3244–3248 (2015)

3249–3253 Physicians’ Acceptance of Electronic Health Records Exchange: An Extension of the with UTAUT2 Model Institutional Trust

Malik Bader. Alazzam, Abd. Samad Hasan Basari, Abdul Samad Sibghatullah, Mohamed Doheir, Noorayisahbe Mohd Yaacob, and Farah Aris

Adv. Sci. Lett. 21, 3249–3253 (2015)

3254–3257 An Integer Programming Model for the Integrated Hub Location and Multi-Hub Vehicle Routing Problem

Ji Ung Sun

Adv. Sci. Lett. 21, 3254–3257 (2015)

3258–3261 A Non-Parametric Measurement of Supercomputers Performance

Corrado lo Storto

Adv. Sci. Lett. 21, 3258–3261 (2015)


3262–3266 Switch Control Scheme to Mitigate Conducted Electromagnetic Interference Emission in Light Emitting Diode Driver

Mohammad Yanuar Hariyawan, Risanuri Hidayat, and Eka Firmansyah

Adv. Sci. Lett. 21, 3262–3266 (2015)

3267–3270 Localization of Index Finger Using Radar Cross Section Measurements for Touchless Keypad Model

Wipassorn Vinicchayakul, Pichaya Supanakoon, and Sathaporn Promwong

Adv. Sci. Lett. 21, 3267–3270 (2015)

3271–3274 Application of Markov’s Normal Algorithm

Guruh Fajar Shidik and Reza Pulungan

Adv. Sci. Lett. 21, 3271–3274 (2015)

3275–3278 Performance Evaluation of Distance Measurement in Biometric Finger Knuckle Print Recognition

Guruh Fajar Shidik, Syafiq Wardani Dausat, Rima Dias Ramadhani, and Fajrian Nur Adnan

Adv. Sci. Lett. 21, 3275–3278 (2015)

3279–3283 Particle Swarm Optimization for Vendor Managed Inventory Control System of Multi Product Multi Constraints

Dwi Ana Ratna Wati, Nur Rachman Dzakiyullah, Chairul Saleh, and Bayu Pebrian Prakoso

Adv. Sci. Lett. 21, 3279–3283 (2015)

3284–3288 Software Reusability in Green Computing

Haryani Haron, Ibraheem Y. Y. Ahmaro, Syed Ahmad Aljunid, and M. Bakri

Adv. Sci. Lett. 21, 3284–3288 (2015)

40

45

50

55

60

65

70

75

80

0.15 1 10 30

CLASS A(QP)

CLASS A(AVG)

CLASS B(QP)

CLASS B(AVG)Lin

e N

ois

e V

olt

age

dBμV

Frequency(MHz)

0°

45°

90°

315°

270°

225°

180°

135°

Index finger

User

TxRx

Radius 50 cm

Average of Accuracy FKP Recognition withPre-processing CLAHE

80.80

94.54

120.00

80.00

100.00

60.00

20.00

0.00

40.00

Euclid...

Manha..

.

Minko

...

Cheby..

.

Chi...

Canbe..

.

Bray..

.

70.90 64.84

95.35

86.46

94.14S

oft

wre

Reu

sab

ility

Ap

pro

ach

es Design patterns

Component-based

development

Applicationframeworks

• Minimize development time• Improving the maintenance• Reduce recurring problems• Increasing the benefits of reuse• Help during implementation

phase

• Balancing the performance• Easy replacement of obsolete

modules• Minimizing time• Minimizing cost• Improving quality

• Enhancing extensibility• Providing mapping• Reducing costs• Improve software quality• Reducing time

• Saving time• Reduce carbon footprint• Cost saving• Promote the use of

videoconferencing

• Minimize the cost• Use self-modifying code• Minimize time

• Reduce power consumption• Practices favor energy-efficient

products• Cost saving• Uses power management

software

• Save cost• Implement single switch

technology• Enhance the processes

• Reduce power consumption• Minimize e-waste• Uses storage virtualization• Save cost

• Recycle technology relatedconsumables

• Minimize e-waste• Reduce carbon footprint• Keeping harmful materials out

of landfills

Telecommuting

AlgorithmicEfficiency

PowerManagement

Voice overInternet Protocol

Virtualization

MaterialsRecycling

Gre

en C

om

pu

tin

g A

pp

roac

hes


3289–3292 Distortion Analysis of Indoor and Outdoor Limit with Biconical Antenna for Ultra Wideband System

Chairak Deepunya and Sathaporn Promwong

Adv. Sci. Lett. 21, 3289–3292 (2015)

3293–3296 Evaluation of VM Selection Policy in Minimizing Cost Energy VM Migration at Dynamic Virtual Machine Consolidation

Guruh Fajar Shidik, Azhari, and Khabib Mustofa

Adv. Sci. Lett. 21, 3293–3296 (2015)

3297–3300 Indoor Radar Cross Section Measurements of Aluminum Hollow Rod for Ultra Wideband Applications

Wipassorn Vinicchayakul, Pichaya Supanakoon, and Sathaporn Promwong

Adv. Sci. Lett. 21, 3297–3300 (2015)

3301–3305 A Comparative Analysis of Stream Data Classifiers and Conventional Classifiers for Anomaly Intrusion Detection

S. Ranjitha Kumari and P. Krishna Kumari

Adv. Sci. Lett. 21, 3301–3305 (2015)

3306–3308 Information Technology in the Accounting

Farida Yerdavletova

Adv. Sci. Lett. 21, 3306–3308 (2015)

3309–3313 Clustering Bilingual Documents Using Various Clustering Linkages Coupled with Different Proximity Measurement Techniques

Rayner Alfred, Leow Ching Leong, Mohd Hanafi Ahmad Hijazi, Joe Henry Obit, and Kim On Chin

Adv. Sci. Lett. 21, 3309–3313 (2015)


3314–3318 A Genetic Algorithm Based Clustering Ensemble Approach to Learning Relational Databases

Rayner Alfred, Gabriel Jong Chiye, Joe Henry Obit, Mohd Hanafi Ahmad Hijazi, Kim On Chin, and HuiKeng Lau

Adv. Sci. Lett. 21, 3314–3318 (2015)

3319–3321 User Generated Content and Internet User’s Brand Engagement

Farzana Quoquab, Zarina Abdul Salam, and Morteza Zeinali

Adv. Sci. Lett. 21, 3319–3321 (2015)

3322–3324 A Comparative Study on Cosine Similarity Algorithm and Vector Space Model Algorithm on Document Searching

Warnia Nengsih

Adv. Sci. Lett. 21, 3322–3324 (2015)

3325–3329 A Cost-Benefit Analysis of Case-Level Radio Frequency Identification Tagging-Based System on Logistics Service

Moon-Soo Kim

Adv. Sci. Lett. 21, 3325–3329 (2015)

3330–3333 Social Media Appropriation in Informatics Reporting—A Systematic Literature Review

Marlita Mat Yusof, Zuraini Ismail, and Nor Zairah Ab. Rahim

Adv. Sci. Lett. 21, 3330–3333 (2015)

RID f1 …2 …3 …… …n …Vector space representation obtained from DARA

Genetic Algorithm-based K-Means Clustering(The first part of the experiment)

K=2 K=3 RID k=2 class0 0 Append each result

back to target table1 0 …

0 1 2 0 …0 1 3 0 …1 2 4 1 …… … … … …… … n n …

Collection

RID k=2 k=3 …1 0 0 …2 0 1 …… … … …n … … …Collection of best clustering result for each k

Genetic Algorithm-based Categorical K-MeansClustering act as Ensembles mechanism

(The second part of the experiment)

K=2 K=3 Append each resultback to target table

RID k=2 class0 0 1 0 …1 1 2 1 …1 2 3 2 …2 2 4 2 …… … … … …… … n n …

UGC WebsiteContentQuality

UGC WebsiteDesignQuality

UGC WebsiteTechnologyQuality

FunctionalValue

BrandEngagement

Document Cluster Tokenizing

FilteringStemming

Indexing

Term Index


3334–3337 Experiences in Instrumented Binary Analysis for Malware

Charles Lim, Darryl Y. Sulistyan, Suryadi, and Kalamullah Ramli

Adv. Sci. Lett. 21, 3334–3337 (2015)

3338–3342 Gamification as an Educational Technology Tool in Engaging and Motivating Students; An Analyses Review

Mageswaran Sanmugam, Norasykin Mohd Zaid, Hasnah Mohamed, Zaleha Abdullah, Baharuddin Aris, and Salihuddin Md Suhadi

Adv. Sci. Lett. 21, 3338–3342 (2015)

3343–3346 Improved Automatic Spell Checker for Malay Blog

Rayner Alfred, Surayaini Bt Basri, Joe Henry Obit, and Zamhar Iswandono Bin Awang Ismail

Adv. Sci. Lett. 21, 3343–3346 (2015)

3347–3351 Big Data, Cloud and Bring Your Own Device: How the Data Protection Law Addresses the Impact of “Datafication”

Sonny Zulhuda, Ida Madieha Abdul Ghani Azmi, and Nashrul Hakiem

Adv. Sci. Lett. 21, 3347–3351 (2015)

3352–3355 A Survey of Value-Based Factors in Software Development

Noor Azura Zakaria, Suhaimi Ibrahim, and Mohd Naz’ri Mahrin

Adv. Sci. Lett. 21, 3352–3355 (2015)

Address previouslyWritten?

BEGIN

Bi–gram changed?

Y

END DUMP

N

Y

List ofwords

obtained

Eliminate Symboland

alphanumerical

Stemmingprocess +new

rules

Misspelled wordidentification and

correction

Exist?

Identificationof repetitive

words

Replace with itscorrectword

Add toreSpelledWord list Yes?

Identificationof Selangorslangwords

Yes?

Identificationof Opposite

wordsYes?

Spellingembodiment

+n-gram

Yes?

Check withrespelled word

dictionary

Yes

Yes

Yes

Yes

Yes

No

No

No

No

No

Start

End

EliminateEnglish word

Tokenization

Eliminatename entity

NeologismIdentification

ACMDatabase

ScopusDatabase

SpringerLinkDatabase

IEEE XploreDatabase

ScienceDirect

Database

WileyDatabase

Identify relevant studies afterskimming the title

Exclude duplicate studies

Identify relevant studies after readingthe abstract

n=286

n=64

Identify relevant studies after detailassessment

n=26

Identify relevant studies from references(snowballing)

n=28

n=347


3356–3359 Service Orientation Initiative Process Towards Enterprise Services Environment

Muhammad Suhaizan Sulong, Azlianor Abdul-Aziz, Andy Koronios, and Jing Gao

Adv. Sci. Lett. 21, 3356–3359 (2015)

3360–3363 A Comprehensive Internet Application Development Process: Web Engineering Consulting Viewpoints

Azlianor Abdul-Aziz, Muhammad Suhaizan Sulong, Andy Koronios, and Jing Gao

Adv. Sci. Lett. 21, 3360–3363 (2015)

3364–3367 Semantic Hybrid Recommender System

Hendrik, Khamidudin Azzakiy, and Aditya Budi Utomo

Adv. Sci. Lett. 21, 3364–3367 (2015)

3368–3372 Exploring Awareness and Perception on Palliative Care—Internet as Source of Knowledge

Noor Azizah Mohamadali

Adv. Sci. Lett. 21, 3368–3372 (2015)

3373–3377 Teachers Acceptance of Frog Virtual Learning Environment (E-Learning): Case Study of Vocational College

Nurbaya Mohd Rosli, Nurazean Maarop, and Ganthan Narayana Samy

Adv. Sci. Lett. 21, 3373–3377 (2015)

3378–3381 Failure Recovery Framework for National Bioinformatics System

Ahmad Shukri Mohd Noor, Emma Ahmad Sirajudin, and Mohd Yazid Mad Saman

Adv. Sci. Lett. 21, 3378–3381 (2015)

3382–3384 A Review on the Developing Algebraic Thinking

Najihahbinti Mustaffa, Zalehabinti Ismail, Zaidatunbinti Tasir, and Mohd Nihra Haruzuan Bin Mohamad Said

Adv. Sci. Lett. 21, 3382–3384 (2015)

PerceivedUsefulness

ComputerSelf Efficacy

PerceivedEase of Use

Intentionto Use

InstructionalDesign

TechnologicalFactor

Convenience

Instructors’Characteristic


3385–3388 Ant Colony Algorithm and Its Application in the Fruit and Vegetable Wholesale Market in Vehicle Scheduling

Wei Xu, Liang Huang, and Hongren Wang

Adv. Sci. Lett. 21, 3385–3388 (2015)

3389–3391 Significance of Preparedness in Flipped Classroom

Azlina A. Rahman, Baharuddin Aris, Mohd Shafie Rosli, Hasnah Mohamed, Zaleha Abdullah, and Norasykin Mohd Zaid

Adv. Sci. Lett. 21, 3389–3391 (2015)

3392–3395 Building a Data Mart Using Single Dimensional Data Store Architecture with Student Subject: Case Study at Muhammadiyah University of Yogyakarta

Fajar Rianda, Asroni, and Ronald Adrian

Adv. Sci. Lett. 21, 3392–3395 (2015)

3396–3399 Contributing Factors of Online Brand Trust in Airline Industry

Nur Atika Jamuary, Mohd Shoki Md Ariff, Hayati Jamaludin, Khalid Ismail, Nawawi Ishak, and Mohd Sawal Abong

Adv. Sci. Lett. 21, 3396–3399 (2015)

3400–3404 Understanding Attitude of Online Shoppers: Integrating Technology and Trust Factors

Li Yuan Hui, Mohd Shoki Md Ariff, Norhayati Zakuan, Norzaidahwati Zaidin, Khalid Ismail, and Nawawi Ishak

Adv. Sci. Lett. 21, 3400–3404 (2015)

Summarize the harvestpropose confusion

Learning by one selfSelf paced

Exhibit communicationResearching cooperatively

Scientific experimentAccomplish Homework

Layout preview

Out

of c

lass

In th

e cl

assr

oom

Word of Mouth

OnlineBrand Trust

Security/Privacy

Perceived Risk

Good Online Experience

Quality of Information

Brand Reputation


3405–3409 Computer-Supported Cooperative Work (CSCW) in Malaysian Homestay Industry

Puteri Noor Ruzanna Abd Aziz, Abu Osman Md Tap, and Murni Mahmud

Adv. Sci. Lett. 21, 3405–3409 (2015)

3410–3417 A Benchmark Feature Selection Framework for Non Communicable Disease Prediction Model

Daniel Hartono Sutanto and Mohd. Khanapi Abd. Ghani

Adv. Sci. Lett. 21, 3410–3417 (2015)

3418–3421 Website Quality and Consumer Attitude of Online Shopping; The Y-Generation Perspective

Yap Soon Jing, Norzaidahwati Zaidin, Mohd Shoki Md. Ariff, Norhayati Zakuan, Khalid Ismail, and Nawawi Ishak

Adv. Sci. Lett. 21, 3418–3421 (2015)

3422–3425 Characteristics of Trustees and Trustors Affecting Consumer Trust in Online Purchasing

Nur Shafiqah Ghazali, Mohd Shoki Md. Ariff, Khalid Ismail, Abdul Halim Ali, Amir Hasan Dawi, and Nawawi Ishak

Adv. Sci. Lett. 21, 3422–3425 (2015)

3426–3429 Exploring the Relationships of Human Behavioral and Perceived Security Aspects Towards E-Assessment Acceptance

Kavitha Thamadharan, Nurazean Maarop, and Ganthan Narayana Samy

Adv. Sci. Lett. 21, 3426–3429 (2015)

Demographics

Personal DetailsComputer Experiences

Past CSCWExperiences

Task ExperimentsRecall Session

CSCW ElementsQuality of Services

(QoS)

Questionnaires

Non-Communicable DiseaseDataset

WBBC WDBC BUPA

LC ECG PID

Feature Selection Technique

Model Validation

10 Fold Cross Validation

Model Evaluation

Confusion MatrixArea Under Curve

(AUC)

Model Benchmark

Classification Algorithm

DTNNSVMNB

wSVM wPCA wRelief wIG wChi

wCFS wUncertain tTest CBWA

testsign Welch MRweight SAM

MRMR RCCW FCBF PAM SVM-RFE

Difference Test Post Hoc Test

Friedman Test Nemenyi Test

Website Quality

Usefulness

Online attitudeof Gen Y

Ease of Use

Entertainment

Complimentary Relationship

PurchaseIntention

ConsumerTrust

PR

PS

SA

PT

Characteristicsof trustors

Characteristicsof trustees

Human Behaviour andInformation System Factors

• Expected Usefulness (EU)• Quality (QL)

Perceived Security Factors• Trust (TR)• Information Security

Knowledge (ISK)• Ethical Behaviour (EB)

ExpectedBehavioural

Intention


ADVANCED SCIENCE LETTERSwww.aspbs.com/science

Aims and Scope:

Advanced Science Letters is a multidisciplinary peer-reviewed journal with a very wide-ranging coverage, consolidates fundamental and applied research activities by publishing proceedings from international scientifi c, technical and medical conferences in all areas of (1) Physical Sciences, (2) Engineer-ing, (3) Biological Sciences/Health Sciences, (4) Medicine, (5) Computer and Information Sciences, (6) Mathematical Sciences, (7) Agriculture Science and Engineering, (8) Geosciences, (9) Energy/Fuels/Environmental/Green Science and Engineering, and (10) Education, Social Sciences, and Public Policies. This journal does not publish general research articles by individual authors.

Research Topics Covered (but not limited to):

Advanced Science Letters deals with Adhesion Science and Technology, Aeronautics Engineering, Aerosol Science and Technology, Aerospace Engineering, Agriculture Engineering, Agriculture Sciences, Anthropol-ogy, Astronomical Sciences, Biochemical engineering, Biochemistry, Bioengineering, Bioinformatics, Biological Sciences, Biomedical Engi-neering, Biomedical Sciences, Biotechnology, Botany, Ceramic Science and Engineering, Cereal Chemistry, Chemical Biology, Chemical Engi-neering, Chemical Engineering, Chemistry, Civil Engineering, Clinical Sciences, Colloid Science, Communication Science, Composites Science, Computer Science, Engineering and Technology, Dairy Science, Device Engineering, Drug Discovery, Earthquake Science, Ecological Sciences, Educational Sciences, Electrical Engineering, Electronics Engineering, Energy Science and Technology, Environmental Engineering, Environ-mental Sciences, Enzyme Science and Engineering, Food Science, For-estry, Fuel Science, Genetics, Geosciences, Health Sciences, Hydrology, Information Technology, Interface Science, Life Sciences, Lubrication Science, Manufacturing Science, Engineering and Technology, Marine Science, Materials Science, Mathematical Sciences, Mechanical Engineer-ing, Medicinal Chemistry, Medicinal Science, Membrane Science, Metal-lurgical Science and Engineering, Meteorology, Microbiology, Minerals Science, Nanoscience, Nanotechnology, Nanoengineering, Nanomedicine, Nanobiology, Neuroscience, Nutrition Science, Oceanography, Optical Engineering, Optical Sciences, Paleontology, Paper Science, Petroleum Science, Petrology, Pharmaceutical Sciences, Pharmacology, Physics, Plant Sciences, Plasma Science and Technology, Polymer Engineering, Polymer Science, Polymer Technology, Powder Technology, Seismology, Sol-Gel Science, Supramolecular Science, Surface Science, Toxicology, Vacuum Science and Technology, Virology, Waste Management, Water Science, Wood Science and Technology, Zoology, Educational Aspects in all these Research Areas, and Selected Conference Special issues on Edu-cation, Social Sciences, Public Policies at the discretion of Editor-in-Chief.

Readership:

The journal is intended for a very broad audience working in all fields of (1) Physical Sciences, (2) Engineering, (3) Biological Sciences/Health Sciences, (4) Medicine, (5) Computer and Information Sciences, (6) Mathematical Sciences, (7) Agriculture Science and Engineering, (8) Geosciences, (9) Energy/Fuels/Environmental/Green Science and Engineering, and (10) Education, Social Sciences, and Public Policies. This journal does not publish general research articles by individual authors.

Manuscript Submission:

Submit Your Manuscript Electronically as a PDF or MS Word file to the Editor-in-Chief or European Editor or Asian Editor or one of the Associ-ate Editors.

Editorial Office:

ADVANCED SCIENCE LETTERSAmerican Scientific Publishers26650 The Old Road, Suite 208Valencia, California 91381-0751, USATel.: (661) 799-7200 Fax: (661) 799-7230E-mail: [email protected]

Books for Review:

Publications should be sent to the Editorial Office.

Abstracting and Indexing:

Chemical Abstracts; Elsevier Bibliographic Databases: Compendex, Scopus; Cambridge Scientific Abstracts: Biological Sciences Abstracts,

Biotechnology and BioEngineering Abstracts, Biotechnology Research Abstracts, Bacteriology Abstracts (Microbiology B), Neurosciences Abstracts, Engineering Research Database, Technology Research Data-base, Environmental Science and Pollution Management.

Referee’s Report:

Please prepare and submit Reviewer’s Report to appropriate Associate Editors accordingly.

Subscription & Advertisement:

American Scientific Publishers26650 The Old Road, Suite 208Valencia, California 91381-0751, USATel.: (661) 799-7200 Fax: (661) 799-7230E-mail: [email protected]

Annual Subscription Rates (Print Edition):

2015, Volume 21Institutional: US$ 3440 (USA), US$ 3440 (Foreign)Postage and handling charges: US$ 300 (USA), US$ 600 (Foreign).The publisher reserves the right to refuse nonqualified subscriptions.

Web Edition:

Advanced Science Letters will be available via internet. For subscription rates to Web Edition, please contact publisher.

Honorable Claim Period:

Claims for issues not received will be honored only if submitted within 60 days of the issue date for subscribers in North America or 120 days for all other subscribers. Send your written request to American Scientific Publishers.

Digital Object Identifier (DOI):

The DOI identification system for digital media has been designed to provide persistent and reliable identification of digital objects. Information on the DOI and its governing body, the International DOI Foundation, can be found at http://www.doi.org. The DOI appears in the lower right of the first page of each article.

Copyright © 2015 American Scientific Publishers, 26650 The Old Road, Suite 208, Valencia, California 91381-0751, USA

All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form whatsoever by any means (electronic, mechanical, recording, photocopying, scanning), or translated into any foreign language or otherwise without the written permission of the Publisher. Only single copies of complete or partial articles may be made for personal and internal research use as allowed by national copyright laws; however, for copying beyond this the copier pays the stated per copy fee through the Copyright Clearance Center, Inc. (CCC). The registered trademarks, names and similar related mate-rials used in this journal are not to be considered unprotected by law.

Although this journal is carefully produced, the authors, editors, and pub-lisher do not guarantee the information and material contained herein to be free of errors. Contributed statements and opinions expressed in Advanced Science Letters (articles, communications, reviews, and research news) are those of the individual contributors and do not necessarily reflect the opinions of American Scientific Publishers, and its Editors assume no responsibility for them. American Scientific Publishers assumes no responsibility or liability whatsoever for any damage or injury, losses, or costs of any kind to person or property that arise due to the use of any materials, instructions, statements, opinions, methods, procedure, infor-mation, advertising materials, or ideas contained herein, negligence or oth-erwise. American Scientific Publishers expressly disclaims any implied warranties of merchantability or suitability for a particular purpose.

Copyright and Reprint Permissions:

Authorization to photocopy is granted by the ASP, provided that the appropriate fee is paid. Prior to photocopying, please contact the Copyright Clearance Center, Customer Service, 222 Rosewood Dr., Danvers, MA 01923, USA; +1 (508) 750-8400. For all other copying such as distribution, promotional, advertising, sale, collective work per-mission, write to: American Scientific Publishers, 26650 The Old Road, Suite 208, Valencia, California 91381-0751, USA.


Copyright © 2015 American Scientific PublishersAll rights reservedPrinted in the United States of America

R E S E A R CH AR T I C L E

Advanced Science Letters

Vol. 21, 2967–2973, 2015

Low Loss Magnetic Design for

Wireless Power Transfer

Heinz Zenkner∗ and Werachet Khan-Ngern

Department of Electrical Engineering, Faculty for Engineering, King Mongkut’s Institute of Technology Ladkrabang,1 Chalongkrung Road, Ladkrabang, Bangkok, 10520, Thailand

An electromagnetically power transmission system to transmit power at a high power density and with a highefficiency under consideration of EMC aspects is proposed. Product requirements like size and weight aretaken into account. The system uses the resonance wave coupling and an impedance adaption for multi-stageconversion is implemented to optimize efficiency and impedance matching. The transferred power is up to 180 Wat a frequency of 6.78 MHz with a distance between 30 mm and 80 mm. The results show that an efficiencyup to 88% can be achieved. The transmitted power density reaches up to 14 W/cm2. It is shown that dueto the evanescent wave resonance coupling the system emits only very little radiation and thus is capable tokeep EMC/EMF regulations. Wireless power transfer applications such as electric vehicle charging and wirelesspower drive for LED lighting through a wall are demonstrated.

Keywords: Wireless Power Transmission, High Efficiency, Power Density, EMC, EMF, Ferrite Material.

1. INTRODUCTIONWireless power transfer is useful where connection cables are

inconvenient, dangerous or even impossible. But the major disad-

vantages of existing systems are1–3 low efficiency, typical <40%,

and low power density as 0.004 W/cm2, large and heavy equip-

ment and only low amount of power transmissible. In most

cases it is only possible to keep conformity according to EMC

and EMF regulations with huge efforts. This research proposes

energy transfer by evanescent wave resonance coupling with high

power density. Advantages are a smaller size of the system com-

ponents together with higher efficiency. Due to the methodology,

biological side effects can be neglected; the system complies

with the International Commission on Non-Ionizing Radiation

Protection (ICNIRP) regulations. The principle of function is the

evanescent wave coupling like described in Refs. [4, 5]: Basi-

cally using a set-up, consisting of two wire coils each with a

diameter of 60 cm, one coil as a transmitter and a second coil

which transforms the energy to the load (a bulb). The system

there is able to transmit power over a distance of two meters. The

coils resonating at 10 MHz use the resonant wave coupling to

avoid that the energy radiates uncontrolled through the air. In the

first demonstration, the researchers showed that the set up can

transfer power with an efficiency of 45%. The particular con-

cept in this research here is the combination of the relative low

operating frequency of 6.78 MHz, a selected ferrite material that

∗Author to whom correspondence should be addressed.

has its resistive contribution of the complex permeability above

the operation frequency and an optimized impedance matching

between all stages and interfaces with an advanced transmitter

inductor concept to achieve a high efficiency.

2. PROPOSED PRINCIPLE OF HIGH POWER

DENSITY WIRELESS TRANSMISSION

SYSTEM2.1. Resonance Coupling System

The resonance coupling effect is based on the evanescent wave

coupling which provides advantages like almost no stray field

to achieve a high efficiency, to fulfill EMF/EMC requirements

and to achieve a higher distance compared to other mecha-

nisms. In the previous work of André Kurs about wireless power

transfer,4�5 investigations have shown that efficiency in the power

transfer rapidly decreases with increasing distance. Therefore it

is necessary to use large coils with a diameter of 60 cm in

order to achieve longer transmission distance. Furthermore, the

impedance and thus the resonance frequency of the receiving coil

is very likely to be influenced by the load. Thus it is necessary

to minimize this effect being independent to supply any kind of

load. The system of this work composes of two main parts: First

part is the transmitter and the second part is the receiver, the

block diagram is shown in Figure 1.

The transmitter, with a resonance circuit operating at

6.78 MHz is supplied by a power signal generator. The receiver

Adv. Sci. Lett. Vol. 21, No. 10, 2015 1936-6612/2015/21/2967/007 doi:10.1166/asl.2015.6451 2967

R ES E A R CH AR T I C L E Adv. Sci. Lett. 21, 2967–2973, 2015

DistancePowersignalgenerator

Trans-mittercircuit

Receivercircuit

Matchingtransfor-mer

Rectifier,Stabilizasion

Overvoltagelimiter

DC/DCconverterLoad

Fig. 1. Transmitter and receiver system set up and block diagram.

module consists of a resonance circuit and a diode bridge rectifier

to convert the RF voltage into a DC voltage. To buffer the energy

against current dips a bulk capacitor stabilization is added, fol-

lowed by a voltage limiter to protect the following circuitry from

over voltage. A DC/DC buck converter provides a regulated out-

put voltage in the range between 5 V and 24 V where any load

can be connected.

2.2. Set Up of the Transmitter and Receiver Circuit

2.2.1. Transmitter Circuit

The transmitter is set up as a series resonance circuit, to provide a

stable behavior of the power amplifier due to the low impedance

at resonance. The transmitter coil has 5 turns, which results in an

inductance of 1.58 �H at a frequency of 6.78 MHz with the cho-

sen ferrite material. The necessary series resonant capacitance,

shown in Figure 2, is in accordance to Eq. (1).

XL = XC

�L= 1/�C

C = 1

L�2= 1

1�58×10−6 · �2� ·6�78×106�2≈ 349pF

(1)

For achievement of a high efficiency it is essential to use a

proper ferrite material6�7 for the inductors and not an air coil.

A prototype of a wireless power transfer system is built in

Ref. [8]. The contribution of the energy in the transmission reso-

nance system must be primarily magnetic. This is achieved by a

high inductance and a low capacitance. The ferrite characteristics

give the advantage to reduce the circuit in size and to increase the

efficiency because of a field concentration and a proper adaption

to the field impedance. The ferrite material has its resistive part

of the complex permeability above 8 MHz and thus keeps the

losses low. The chart of the permeability is shown in Figure 3.

The material is selected in such a manner that at the operating

frequency the imaginary contribution (u′′s ) of the complex perme-

ability is almost zero what keeps the losses inside the ferrite at a

minimum.9

Transmission InduktorL = 1.58uH

Power Input

C = 349 pF – Cx/2 Cx: Variable capacitorfor fine-tuning

Fig. 2. Transmitter resonance circuit.

us`= 125

us``~ 1 at 6.78 MHz

Resistive(losses)

Reactive

Fig. 3. Ferrite characteristic: Complex permeability as a function offrequency.7

The complex permeability represents the “magnetic behavior”

of the ferrite material and has two contributions, the reactive

portion (u′s), which represents the inductance shown in Eq. (2),

and the resistive portion (u′′s ), which represents the losses shown

in Eq. (3). The initial permeability of the ferrite material is 125

as shown in the diagram.

LS = Lo ·u′s (2)

Rs = ��L� ·u′′s (3)

Then, the total impedance is

Z = �j�L� · �u′s −u′′

s � (4)

The absorption factor can be defined as

tan= Rs/�Ls = �u′′s �/�u

′s� (5)

With implementation of a ferrite core the losses in terms of

a resistive component add and the simplified diagram, without

parasitic effects is shown in Figure 4.

In case of an alternating magnetization of the ferrite the flux

density B is not in phase with the magnetic field produced. In

case of “small” magnetization, the angle between the result-

ing magnetization and the flux density represents the loss angle

shown in Eq. (5) and Figure 5.

The smaller the angle the lower are the losses and the higher

is “quality” of the material.9�10 The over-all loss angle of the

inductance is a combination of the contribution of the ferrite and

the wires of the coil. In case of larger magnetization the main

losses of the ferrite are the hysteresis, eddy current and magnetic

creep losses. The high specific resistance of material, which is

NiZn-ferrite shown in Table I, is required to reduce the eddy

currents losses.

Fig. 4. Ferrite core loss components, stray capacitances neglected.

2968

R E S E A R CH A R T I C L EAdv. Sci. Lett. 21, 2967–2973, 2015

Fig. 5. Ferrite core loss parameters.

With rising frequency, current does not flow homogeneous

through the entire cross-section of the conductor but is more con-

centrated at the surface. The higher the frequency, the more cur-

rent is concentrated on the surface. This results in higher I2 ∗Rlosses and thus energy loss with rising frequency, the current den-

sity varies exponentially as a function of depth from the surface

of the wire. The effect is called skin effect. The skin depth is

defined as the distance below the surface where the current den-

sity has fallen to 1/e or 37% of its value at the surface.9 The skin

depth at 6.78 MHz for a copper wire is calculated and expressed

in Eq. (6).

=√2/��= 1√

�f�0�r(6)

�: Angular frequency of current (2�f in 1/s)

�: Magnetic permeability (�r ·�o� : Conductivity of the material (S/m or 1/�m).

= 1�05 1/�m

�0 = 4� ·10−7 H/m

cu = 5�82 ·107 �/m

=√

2

2� ·6�78 ·106 ·4� ·10−7 ·5�82 ·107 ·1�05= 24�7 �m

The parameter above show possible high losses due to the

skin effect. A similar effect, the Eddy current losses in the wind-

ings cause a main contribution of the losses. They are depending

on following parameters of Eq. (7), whereas the material and

the field must be uniform and the skin effect may here not be

considered:11

P = �2B2pd

2f 2

6k�D(7)

P : Power lost per unit mass (W/kg),

Bp: Peak of the magnetic field (T),

D: Diameter of the wire (m),

F : Frequency (Hz),

k: A constant equal to 1 for a thin sheet and 2 for a thin wire,

�: Resistivity of the material (� m), and

D: Density of the material (kg/m3).

This equation is valid only under the so-called quasi-static con-

ditions, where the frequency of magnetization does not result in

the skin effect what means the electromagnetic wave fully pen-

etrates the material which is the case at very low frequencies.

Table I. Specific resistance of selected materials.

Composition Specific resistance at T = 25 �C (�m)

MnZn-Ferrite 0.1–10NiZn-Ferrite 105–106

Ferrite

Strands, twisted

Fig. 6. Litz wire of the transmitter ferrite core.

But it shows clearly the dependency of the wire parameters like

thickness (d) and material constants (� and D) which certainly

do influence in case of higher frequencies, too. The other param-

eters are system depending, cannot be varied and thus not be

optimized.

Another factor, which may influence the efficiency, is the cur-

rent density, which is the ratio of current intensity to the area,

perpendicular to current direction, through which the current is

flowing. The mathematical definition of current density, which is

applicable to any possible distribution of charges flowing in the

conductor is11

I =∫s

�J ·d �S (8)

where �J is the current density at the area element d �S, and I is

the total current through area [A/m2]. In case one wire/strand

with a diameter of � 0.8 mm is used for the transmission coil,

a current density of up to 65�4∗106 A/m2 is caused, which is far

too much. The current density limitation for standard applications

is 1�0∗106 A/m2, which is about 65 times less! The construction

method of the transmitter coil is essential for the high efficiency

of the system. The assembling of a coil is depending on several

parameters: Number of turns, material of wire, diameter of wire

and for the litz wire the number of wires and the method of

twisting. The following picture (Fig. 6) shows the litz wire on

the core, which is eventually used.

Another important parameter of the transmission resonance

system is the self-resonant frequency of the coil which is a

parallel resonance consisting of the inductance and the parasitic

capacitances between the windings of the coil. As the series

capacitance of the resonance circuit is quite small, the impact

of the parallel capacitance must be considered and kept small.

C1

111 pF

10:3

L2

P SL3

Fig. 7. Receiver resonance circuit with decoupling transformer.

2969


It is necessary to use a suitable wire size and a proper layout.11

As mentioned in terms of the current density the variation of

the coil may change with the number of turns, twisting fac-

tor, amount of litz wires, density of turn layout, the impedance

matching to power amplifier and to the system receiver and the

magnetic field contribution of the efficiency.

2.2.2. Set Up of the Receiver Circuit

The receiver resonance circuit has to supply a nonlinear load

like the rectifier bridge, bulk capacitors and the DC/DC converter

and has to match to the near field wave impedance conditions.

Therefore, a parallel resonance circuit has been chosen. For opti-

mizing the decoupling between the resonance circuit and rectifier

bridge and to increase the efficiency by impedance matching a

decoupling transformer has been implemented (Fig. 7).

The parameter of the inductance and the capacitance can be

calculated based on the same resonant frequency, 6.78 MHz

Power meter Input / Output wave

Power meterheads

Poweramplifier

Signalgenerator

Test set-up measuring system

Test set-up diagram

Dirctionalcoupler

Voltage/Current meters

Transmitter,Receiver

Receiver

Transmitter

(a)

(b)

Fig. 8. Test set-up for the precise measurements of the system efficiency.

Reflected signal from load (Channel B): 130 mVss

Signal from Generator(Channel A):920 mVss

Fig. 9. Signals at directional coupler (50 ns/div, 200 mV/div).

but with consideration of the transformed additional impedances.

With the implementation of the additional decoupling, resp.

matching transformer the AC/DC power for any nonlinear load

can be supplied.

3. TRANSMISSION PERFORMANCE OF THE

SYSTEMThe centered diameter of the core is 42 mm, thus the efficiency

is measured with a gap of 42 mm as set-up in Figure 8. Figure 9

illustrates the signals decoupled by a 40 dB directional coupler

at the oscilloscope.

The upper trace is the reflected voltage from load at output B

(Fig. 8(a)), about 130 mV and the lower trace is the signal from

preamplifier at output B with 920 mV at an operating frequency

of 6.78 MHz (50 ns/div).

Result:

� After warm up (>1 h): core: 29 �C, coil: 32 �C,

Pin: 10.2 W, VDC: 22.5 V, RL: 56 Ohm

�% = �22�5�2

10�2×56·100 ≈ 88�6%

The transmitted power density is defined as the ratio between

the maximum transmitted power and the active area of the induc-

tor core (Fig. 10). At a transmitted power of 80 W with the

cross section of the ferrite core like shown in Figure 10 which

is 15�3× �102�4− 65�5� = 564�57 mm2 or 5.65 cm2 the maxi-

mum transmitted power density achieved is 0.14 W/mm2 which

is 140 kW/m2! For comparison, the power density of the sun

light is 1.37 kW/m2, the power density of Uran12 is 650 kW/m2.

4. SYSTEM LOSSES AND EMC/EMF

CONSIDERATIONSSystem losses are uncontrolled radiation and heat conversion.

The losses in the system compose of losses in the inductor (inner

and outer), losses in the capacitors and losses in the DC conver-

sion components. Some of those effects sum up and the energy is

converted into heat, others may radiate. The main contribution of

102.4 mm

65.5 mm

15.3 mm

Fig. 10. Cross section area of the transmitting and receiving core.

2970


Lowest level:0.1uT

–

Operating Frequency:6.78MHz

Fig. 11. ICNIRP limits of magnetic flux density exposure to human body.

the losses in the circuit are those of the inductors, the transmit-

ter capacitor is the second large contribution of the losses. The

losses of the inductors are contributions of the ferrite material

and of the wire windings. The primary losses of the ferrite are the

hysteresis-, eddy current- and magnetic creep-losses, the losses

of the wire windings are resistive losses and eddy current losses

due to the skin- and the proximity effects.11 The proper selec-

tion of the ferrite material and the optimal construction of the

wire winding is essential to reduce the losses and thus to reach

an optimum of efficiency.13 The losses in the set up are mainly

impedance matching losses between the power transmitter and

the transmission resonance circuit.

For keeping the EMC requirements the CISPR 1114 standard

for industrial, scientific and medical (ISM) equipment is consid-

ered. CISPR 11 defines ISM-designated frequency bands, which

are exempted from emission requirements. This means that there

are no radiation limits at certain frequency ranges. One range

defined in CISPR 11 is from 6.765 MHz to 6.795 MHz with a

center frequency of 6.780 MHz.

Today it is a must for a state of the art product to consider the

biological effect on human body. The International Commission

on Non-Ionizing Radiation Protection (ICNIRP) publishes guide-

lines for limiting RF exposure that provides protection against

bigcore

Strayfield

Sensors

Fig. 12. Set up of radiation measurement with a high sensitivity sensor.

Spectrumanalyzer

Field probes

Transmission systemRF source

Loopantenna

Powermeters

Load

Fig. 13. Set up of the wireless power transmission system to evaluate theEMC radiation.

known adverse health effects.15 In this work it can be shown

that the field used to transfer the power is limited to the area

between transmitter and receiver and thus outside of the beam

has less biological effect. The set-up can even keep the limit level

of 0.1 �T , which is the lowest level in the ICNIRP limit chart

starting at a frequency of 10 MHz and going up to 300 MHz as

shown in Figure 11.15

With a high sensitivity sensor, detecting fields in the range of

0.05 uT, it can be shown that the field of the inductance is limited

to the area between transmitter and receiver and only susceptible

to receivers at the same resonance frequency and thus outside of

the beam has no considerable biological effect. Figure 12 illus-

trates the set-up.

For further investigation, the system is set up like shown in

Figure 13. The RF source, an amplifier of 20 W is put in a

shielded box to avoid any unwanted radiation. The transmitter

and the receiver are placed on a non-metallic box which has no

influence on the electromagnetic behavior of the system. At a dis-

tance of 1.2 m a magnetic loop antenna is placed and connected

Power beam

Transmitter coil Receiver coil

Center

Fig. 14. Measurement points where H-field and E-field emission iscaptured.

2971


Table II. Result of the emission measurements according to the set

up shown in Figure 13.

Level in Level inLoad Kind of Level at Level at power center of Antennacondition field transm. Rec. beam core level

High (15 W) electric 102 90 102 102 67Low (5 W) electric 102 91 102 102 67High (15 W) magnetic 74 78 116 84 67Low (5 W) magnetic 82 80 94 85 67

Note: All results in dBuV (relative).

to a spectrum analyzer to measure the magnetic field emission of

the system in dependence of different operating conditions.

At close distance to the transmitter and the receiver the emis-

sion of the electric field and the magnetic field is measured

with field probes. The measurements are done in the vicinity of

the ferrite cores, in the middle of the core system and directly

between the poles (Fig. 14).

The emission is in fact, besides of the heat losses, another

contribution of system losses, which decreases the efficiency of

the system. The radiation losses can be separated into classical

losses, which are those, radiated from cables and components

because of mismatch and parasitic impedances and in losses,

which occur because of the power transmission beam and actu-

ally should be, due to the nature of the resonance evanescent

wave-coupling, zero. Table II lists the result of the measurements.

Two load conditions are measured, one with 5 W power con-

sumption another with 15 W. At both load conditions the electric

and the magnetic field are measured at the vicinity of the trans-

mitter and the receiver, in the power beam and at the center of

the transmission core. Additionally the radiation of the magnetic

field is measured with a magnetic loop antenna. Figure 15 shows

the signal received by the loop antenna in low power operating

condition.

The results listed in Table II allow following important

conclusions:

(1) The electric field is at the transmitter very high due to the

high operating voltage at the capacitor. It surrounds the whole

transmission system with the same high level.

(2) The magnetic field is highest in the power beam and

decreases rapidly in the center of the cores. There is also a low

level surrounding the cores caused by stray fields of the coils.

Fig. 15. Emitted signal measured with the loop antenna, operating condi-tion: 5 W load at the receiver.

Fig. 16. Transmitting the power through the cement wall, 9 cm thickness.

(3) The magnetic field measured with the Loop antenna is inde-

pendent from the load (!). It shows that the power transmission

is restricted to the area between the cores.

The remaining emission is mainly caused by

—Radiation from the cable between the power amplifier and the

transmitter, due to mismatch and endless shielding effectiveness.

—Parasitic impedances of the transmitter and receiver resonance

circuits.

—Proximity and skin effect of the coils.

In conclusion when the receiver is taken away from the sys-

tem the remaining emission level at the loop antenna decreases

only to approx. 65 dBuV. Due to the high quality factor Q of

the transmitter resonance circuit the emission of any harmonics

caused by the RF power amplifier is very low what gives the

possibility to implement a hybrid class D RF-amplifier in small

size as a next application.

5. APPLICATIONSThe application in Figure 16 shows a wireless power transmission

through a 9 cm cement wall.

A further application may also be for example a storage energy

charging system for cars or in medical appliances. Applications

of other researchers focus on areas such as: Wireless power

transmission with multi receivers in power supply system’s,16

wireless power and data link,17 wireless charging systems18 or

wireless power standardization for supply and charging of small

appliance.19 Each of the applications has its individual advantage

but none of them can hardly combine the key features which

are small size, proper distance between transmitter and receiver,

EMC consideration and high efficiency. The applications operate

in the range from 100 kHz up to 27 MHz so many not in the

ISM band, some reach an efficiency up to 90% but work at low

distance and on magnetic coupling basis only.

6. CONCLUSIONIn this work it is shown that with pre-defined parameters and

restrictions for practical use as a precondition, it is possible to

achieve a result, which can be used already for industrial design.

The work shows that it is possible to transmit high power via

the air keeping still a high efficiency, a low weight, a small size

and also EMC/EMF restrictions. The circuit enables a power

2972


transmission under conditions, which make the receiver safely in

use as no cables, or batteries need to be applied. Thus there is a

wide range of application possible like in medical appliance or

appliance for chemical industry where special precautions must

be taken.

The parameters of wireless power performance are the distance

between transmitter and receiver, the transmitting frequency and

the system impedances. The key for the principle is the resonance

coupling effect together with a high performance ferrite inductor

which allow to fulfill the preconditions like EMC/EMF and effi-

ciency. The various load conditions normally affect the matching

impedances and thus they have to be decoupled from the transmis-

sion system here realized with implementation of an additional

matching transformer with again a proper high efficiency ferrite

material. The efficiency is significantly depending on the trans-

mitter resonance circuit where the highest power of the whole

system is handled. The components used here have to be opti-

mized regarding low loss and small size. The maximum efficiency

of 88.6% is achieved at a gap of 42 mm with 9 W output power.

The maximum transmitted power density achieved is 0.14 W/mm2

which is 140 kW/m2! The component and material selection and

design are essential to reach a high performance transmission. The

further progress in this work is focused on how to maintain the

high efficiency wireless power transmission at various distances

and how to keep the high efficiency at longer distance.

References and Notes1. J. J. Dubray, Standards for a service oriented architecture (2003), http://

www.ebxmlforum.org/articles/eb for_20031109.html.2. D. H. Akehurst, Transformations based on relations (2004), http://heim.

ifi.uio.no/_janoa/wmdd2004/papers/akehurst.pdf.

3. A. Kurs, A. Karalis, R. Moffatt, J. D. Joannopoulos, P. Fisher, and M. Soljacic,Science 317, 83 (2007).

4. A. Karalis, J. D. Joannopoulos, and M. Soljac, Elesevier, Annals of Physics323, 34 (2008).

5. A. Karalis, J. D. Joannopoulos, and M. Sol-jacic, Elsevier Annals of Physics323, 34 (2008).

6. H. Zenkner, A. Gerfer and B. Rall, Trilogy on Inductors, 3rd edn., Swiridof-fVerlag (2009), pp. 140–143, ISBN: 3-934350-73-9.

7. Ferroxcube: Soft Ferrites and Accessories, http://www.ferroxcube.com, DataHandbook (2008), Vol. 29, pp. 7–13.

8. B. L. Cannon, J. F. Hoburg, D. D. Stancil, and S. C. Goldstein, IEEE Transac-tions on Power Electronics 24, 1819 (2009).

9. Meinke, Gundlach, Taschenbuch der HF-Technik Bände I–III, Hrsg. Von K.Lange, and K.-H. Löcherer, Springer-Verlag (1992), pp. B3–B4, E13–E14,H1–H4, 17–19, ISBN: 3-540-54714-2.

10. K. G. Kaschke and H. Gmb, Co. Rudolf-Winkel-Str. 6, 37079 Göttin-gen, Nickel-Zink-Kobalt-Ferrite, http://www.kaschke.de/fileadmin/user_upload/documents/datenblaetter/Materialien/NiZn-Ferrit/K251.pdf, datasheet.

11. H. Kaden, Wirbelströme und Schirmung in der Nachrichtentechnik, edited byH. W. Meissner, Springer-Verlag (1959), Vol. 111, pp. 59–88.

12. Energy density at WWW.Wikipedia.org/wiki/Sonnenenergie.13. H. Zenkner, A. Gerfer, and B. Rall, Trilogy on Inductors, 3rd edn., Swiridoff-

Verlag, pp. 60–81, ISBN: 3-934350-73-9.14. CISPR 11:—Industrial, Scientific and Medical (ISM) Radio-Frequency

Equipment—EM Dis. Char.—Limits and Methods of Measurement. CISPR11,IEC (2011).

15. ICNIRP Guidelines: Guidelines for limiting exposure to time-varying elec-tric, magnetic and electromagnetic fields (up to 300 GHz), http://www.icnirp.de/documents/emfgdl.pdf, p. 512.

16. TinekeThio, A Bright Future for Subwavelength Light Sources, American Sci-entist (2006), Vol. 94, pp. 40–47.

17. R. Puers, K. V. Schuylenberght, M. Catrysse, and B. Hermans, Wirelessinductive transfer of power and data, Analog Circuit Design, Springer (2006),pp. 395–414.

18. R. Hui, Comparison of Power Savings Based on the Use of Wireless ChargingSystems and Conventional Wired Power Adapters, City University of HongKong (2009).

19. Wireless Power Consortium, Transfer efficiency, http://www.wirelesspowerconsortium.com/technology.

Received: 9 October 2014. Accepted: 19 November 2014.

2973




Vol. 21, 2997–3001, 2015

Computational Modeling of Mood from

Sequence of Emotions

Dini Handayani1�∗, Hamwira Yaacob1, Abdul Wahab Abdul Rahman1,Wahju Sediono2, and Asadullah Shah1

1Department of Computer Science, Kuliyyah of Information Communication and Technology2Department of Mechatronic, Kuliyyah of Engineering, International Islamic University Malaysia Kuala Lumpur, Malaysia

In the recent years, more studies that aim to make computers understand, experience and respond to variousemotional states accordingly through computational models have been widely researched. Conversely, little hasbeen done to recognize the medium term of emotion, such as mood. Thus, in this study, a mood modeling wasproposed to recognize mood from a sequence of several emotional states. The input for the proposed moodmodel was derived in the form of electroencephalogram (EEG) signals, which were captured from five subjectsduring eyes close, and eyes open. Our analysis indicates that mood can be recognized either from eyes closeor from eyes open.

Keywords: Computational Modeling, Mood, Emotions, Affective Computing.

1. INTRODUCTIONIn the field of affective computingm,1 recognition of emotional

states has been widely studied in several domains, including

education,2�3 health,4 and entertainment.5 Results from the recog-

nition of emotional states may be used to analyze user reactions.

Accordingly, the user intention can be predict and the appropriate

response can be produce.

In general, the words mood and emotion are used interchange-

ably. Emotion is driven by specific events, actions or objects,6

while mood is not related to a specific event.6 Emotion is defined

as a short-term effect or punctual emotional state that occurs very

brief in duration,7 whereas mood reflects as the medium term

affect with the duration longer than emotion.6

Mood is a collection of several emotional states in a certain

duration of time, and it carries information regarding the environ-

ment, and usually not distinguishable through facial expressions.8

On the other hand, emotion is accompanied by specific facial

expression.9 Emotion and mood can mutually influence each

other. Emotion, if it is strong and deep enough, will turn into

mood.

The aim of this research is to construct a mood computational

model from a sequence of emotional states, to analyze the tem-

poral properties of the model and to propose the model for auto-

mated mood recognition.

Among others, technology endowed with this skill can drive

or maintain a person positive mood. People in positive mood


make good decisions quickly, enhance social relationships, pro-

duce more ideas, and tend to identify more creative options to

problems.3�7�8

This paper is organized as follows. In Section two, the related

work is presented. Methodology is then elaborated in Section

three. Experimental result is discussed in Section four. Finally,

our work is summarized in the last section.

2. RELATED WORKPrevious studies indicate that a computational model of moodhas been evaluated based on either implicit or explicit approach.

Through an implicit approach, mood information is inferred

based on the measurement of the behavioral and physiological

signal of the user. Some of the implicit mechanisms that are

used for measuring mood from different physiological conditions

include body gesture and facial expression.10–12 On the other

hand, the explicit measurement is done by direct self-assessment

of user emotional states.13�14

According to discrete view of emotion, some researcher defines

emotion as sad, happy and neutral emotion.11�12 Some researchers

believe that mood can be described based on two affective dimen-

sions, namely valence and arousal.13�15 Others considered three

affective dimensions in their studies.10�14

Most of research works were used classification techniques

for mood formulation.10–12 Two studies were used statistical

approach for mood formulation, such as average14 and moving

average (MA).13



Table I. Taxonomical table on mood formulation.

No Author Measurement tools and data Emotion recognition Mood formulation

1 Katsimerou et al.13 Self assessment, audio-visual TheHUMAINE affective database

Dimension: Valence and arousal Moving average

2 Thrasher et al.10 Body gesture during listening music Dimension: Valence, energetic arousal,and calmness.

Logistic regression from bodygesture

3 Livne et al.11 Data of 3D pose tracking and motioncapturing

Discrete: Sad and happy Logistic regression videoanalysis

4 Metallinou and Narayanan14 Self assessment, audio and video fromUSC CreativeIT database

Dimension: Activation, valence anddominance

Average of video analysis

5 Hashemian, et al.12 Face reader, interaction with computer Discrete: Sad, happy and neutral Bayes classification

Based on our literature review as shown in Table I, some

research works were implement to recognize mood from video.

Three studies reported on how mood of a real subject is directly

recognized in a real-time.10–12 None of the study utilize electroen-

cephalogram (EEG) for mood recognizer. Hence, this motivates

us to explore more on this matter.

In this study, a computational model of mood were constructed

from a sequence of several emotional states, its temporal prop-

erties were analyzed and proposed it as a model to automatic

recognize the mood from EEG.

EEG is an imaging tool that captures electrical activation

occurrences which is used to monitor the brain condition.16

Brainwaves which are represented by EEG signals are commonly

ranged into four bands; delta (0.5 Hz–4 Hz) Characteristic of

deep sleep phases,17 theta (4 Hz–8 Hz) Drowsiness and fatigue

due to monotonous task,18 control of working memory process,19

alpha (8 Hz–13 Hz) Cognitive control,20 creative thinking21 and

beta (13 Hz–30 Hz) Alertness,22 phonological tasks.23 Each of

the frequency bands was observed as products of different brain

tasks.

3. METHODOLOGYThe methodology of this research will be done with the following

step:

3.1. Signal Acquisition

Four EEG electrodes (C3, C4, T3, and T4) were pasted on their

scalp, with the specific regions using the International 10–20 sys-

tem. The electrodes will be connected into the EEG head box to

enhance the signals.

3.2. Signal Preprocessing

Here, all the signals were filtered to exclude noises and unre-

lated artifacts. Finally, only one-minute activities in the recorded

signals were used for the training.

Fig. 1. Flowchart of the methodology.

3.3. Feature Extraction

Kernel Density Estimation (KDE) was applied as a features

extraction technique. 100 features were extracted from the signals

for each instance.

3.4. Classification

Multi Layer Perceptron (MLP) was selected for classifiers. The

training was done based on dimensional approach. The classes

were labeled based on the generalized values of valence and

arousal. The values were assigned based on quadrants in which

each emotion is located in the affective space model, as shown

in Table II.

Bialoskorski et al.,24 defined color for emotional state. Happy

emotional state is indicated as having positive level of valence

and high arousal with an orange color. Quadrant of positive

valence and low arousal represents the calm emotional state with

green color. Sad emotion is located at the quadrant of negative

valence and low arousal with blue color. Finally, fear emotion

corresponds to negative valence and high arousal with red color.

During the experiment, the data obtained from the subjects

were fear, sad, and happy emotion.

For the dimensional approach, the performances are consis-

tent. For valence dimension, subject three gets the highest with

94.93% accuracy. Subject two and subject four get the lowest

with 90.92% accuracy. With the mean (M) 92.04% and standard

deviation (SD) 1.67.

For arousal dimension, subject three gets the highest with

93.36% accuracy. Subject two get the lowest with 90.92% accu-

racy. With M 92.18% and SD 1.31. As shown in Figure 3.

Figure 4 shows the scatter plots of the basic emotional states.

It is distributed well within the expected quadrants as described

in ESM.

3.5. Mood Formulation

Subjects are instructed to do the resting states (eyes open and

eyes close), one minute for each condition. Two episode of an

emotional state of a user were collected with a total time for

each episode is one minute (n). Sample rate for each emo-

tions are 0.004 seconds (t). One emotion can be estimated from

Table II. Valence and arousal labels for emotions classification.

Labels

Emotions Valence Arousal

Happy 1 1Calm 1 −1Fear −1 1Sad −1 −1

2998


Fig. 2. The emotional states model (ESM).

the sample rate. However, the overall episode will refer to the

moods.

In this research both emotions and mood refer to emotional

state of the EEG signal from the subject. From the classifi-

cation phase, basic emotion and label of the emotions were

defined.

e= �v�a� (1)

For the every episode, the emotion vector e, corresponding to

the recognize time as follow:

et = �vt�at�� t = 1�2�3� � � � � n (2)

Assuming E as sequence emotions

E = �e1� e2� e3� � � � � en� (3)

Finally, mood formulation is measured based on the corre-

sponding mood function, as a proposed work from the previous

study.25

m= F �E� (4)

The proposed mood model is based on MA of emotions over

time. The mood formulation, were divide into valence and arousal

mood.

M = �Vm�Am� (5)

For mood of valence, the formulation as follows:

Vm = MA of Valence (6)

Vm = Vt+1 (7)

Vt+1 = Vt−1 +��Vt −Vt−1� (8)

Table III. Parameters for MLP.

Parameters Values

Number of hidden layer 1Number of nodes in hidden layer 30Mean-square error goal 0.1Activation function at hidden layer Tan-sigmoidActivation function at output layer Pure linier

Fig. 3. Accuracy of Subject’s Identification.

Where: Vt+1 = Valence for the next period, �= Smoothing con-

stant, Vt =Observed value of valence in period t, Vt−1 = Previous

valence.

Likewise, Eqs. (6)–(8) can be re-written for arousal as:

Am = MA of Arousal (9)

Am = At+1 (10)

At+1 = At−1 +��At −At−1� (11)

Where: At+1 = Arousal for the next period, �= Smoothing con-

stant, At =Observed value of arousal in period t, At−1 = Previous

arousal.

3.6. Mood Recognition

Mood recognition is derived from the most prominent emotion.

Thus, the individual emotion was map into mood directly and

takes the quadrant of the mood space that containing the major-

ity of emotion. This quadrant may then be a predictor of the

recognized mood.

Fig. 4. Emotions distribution for subject 3.

2999


Fig. 5. Subject 2 eyes open emotions.

Fig. 6. Subject 2 eyes open, mood of the emotions.

4. EXPERIMENTAL RESULTThe main interest of this research will be the correlation between

emotion and mood of the subject. In addition, it is important to

get a higher accuracy, to prove that using dimensional approach

(valence and arousal) are better than categorical approach, as

discussed in Ref. [16].

Thus, memory test for all subjects were constructed to see the

level of accuracy, either it can be accepted or rejected as shown

in Figure 3. Eyes close and eyes open signal will be compared

to get mood recognition result.

The eyes open emotional states for subject two are shown in

Figure 5. The subject has the dynamic emotion. Initially, the sub-

ject was in the sad emotion, change to fear and happy emotions.

Calm emotions also appear during the eyes close.

The get the mood of emotion, MA were calculated for both

valence and arousal. With the smoothing average is 1/60 seconds.

As depicted on the Figure 5, the shortest emotions is one second

from the total episode 60 seconds.

�= 1

60sec

As shown in Figure 6, the emotions of subject two were

changed. The prominent emotions are sad emotions. Here, the

mood states can be derived from the prominent emotion. How-

ever, calm and fear emotion were still appear. To get the most

prominent emotion, MA was recalculated from the mood value.

From the Figure 7 above, the most prominent emotion is sad.

Therefore the mood of subject two in eyes open is sad mood.

Fig. 7. Subject 2 eyes open, mood of the mood.

Table IV. Experimental results.

No Subject Action Emotion Mood 1 Mood 2

1 Subject 2 EC – Sad Sad

2 Subject 3 EC – Fear FearEO – Sad Sad

3 Subject 4 EC – Happy HappyEO – Happy Happy

4 Subject 5 EC – Calm HappyEO – Calm Calm

5 Subject 6 EC – Happy HappyEO – Calm Calm

EO – Sad Sad

The mood from eyes open will be compared to eyes close. With

the assumption, the mood recognition will be the same.

As a summary, we have Table IV as an experimental result.

It is consist of five subjects with eyes close and eyes open. The

moods of eyes close and eyes open are similar for subject two

and four. However, for subject three, five and six, the mood is

different. Even though the emotions are different, all the emo-

tions are in the same valence area. Therefore, subject three in the

negative mood, subject five and six are in the positive moods.

5. CONCLUSIONSIn this paper, computational models of mood that infer the long-

term emotional state of a person have been proposed. The emo-

tional state is generated from EEG brainwave. Two emotional

episodes are taken into consideration; (1) eyes close and (2) eyes

open. It is showed that a model could get the prominent emotions

from second order of moving average as a mood recognition for-

mula. For the future work, mood recognition will be under the

executive function task, and correlate the result with the resting

states.

In addition, it is expected that the refined models are being

able to properly capture the process that regulate the relationship

between recognized emotions and mood. Finally, valence and

arousal values are predicted quite satisfactorily from the proposed

models, yet their combination into mood is less precise.

Acknowledgments: We would like to express our sincere

gratitude to Norzaliza M. Nor for providing us with the data

and valuable feedback. This work is supported by Fundamen-

tal Research Grant Scheme (FRGS) funded by the Ministry of

Higher Education (Grant code: FRGS14-137-0378).

References and Notes1. R. W. Picard, Int. J. Hum. Comput. Stud. 59, 55 (2003).2. D. Papachristos, K. Alafodimos, and N. Nikitakos, Emotion Evaluation of Sim-

ulation Systems in Educational Practice (2012), pp. 1–7.3. C. Pimentel, Affect. Comput. Intell. Interact. 72 (2011).4. S. Alghowinem, R. Goecke, M. Wagner, G. Parkerx, and M. Breakspear,

Head pose and movement analysis as an indicator of depression, 2013Human Association Conference Affective Computing and Intelligent Interac-tion, Sepetember (2013), pp. 283–288.

5. B. Bostan, Entertain. Comput. 262 (2010), no. Idmi.6. I. Siegert, R. Böck, and A. Wendemuth, Cogn. Behav. Syst. 273 (2012).7. J. H. Janssen, E. L. V. D. Broek, and J. H. D. M. Westerink, User Model.

User-adapt. Interact. 22, 255 (2011).8. D. Hume, Organ. Behav. 258 (2012).9. F. D. L. T. Torre and J. F. Cohn, Handb. Face Recognit. 1 (2011).

3000


10. M. Thrasher, M. D. V. D. Zwaag, N. Bianchi-Berthouze, and J. H. D. M.Westerink, Mood Recognition Based on Upper Body Posture and MovementFeatures, Springer-Verlag, Berlin, Heidelb. (2011), pp. 377–386.

11. M. Livne, L. Sigal, N. F. Troje, and D. J. Fleet, Comput. Vis. Image Underst.116, 648 (2012).

12. M. Hashemian, A. Nikoukaran, H. Moradi, M. S. Mirian, and M. Tehrani-Doost,Determining mood using emotional features, 7’th Int. Symp. Telecommun.,September (2014), pp. 418–423.

13. C. Katsimerou, J. Redi, and I. Heynderickx, A computational model for moodrecognition, 22nd Int. Conf. UMAP 2014, Aalborg, Denmark (2014), Vol. 8538,pp. 122–133.

14. A. Metallinou and S. Narayanan, Annotation and processing of continuousemotional attributes: Challenges and opportunities, Proc. 2nd Int. Work. Emot.Represent. Anal. Synth. Contin. Tome Sp. (Emosp. 2013), Shanghai, China(2013).

15. H. Yaacob, I. Karim, A. Wahab, and N. Kamaruddin, Two dimensional affectivestate distribution of the brain under emotion stimuli (2012), pp. 6052–6055.

16. H. Yaacob, Classification of EEG signals using MLP based on categoricaland dimensional perceptions of emotions, 2013 5th Int. Conf. Inf. Commun.Technol. Muslim World, March (2013), pp. 1–6.

17. K. Šušmáková, Slovak Academy of Sciences 4, 4 (2004).18. B. T. Jap, S. Lal, P. Fischer, and E. Bekiaris, Expert Syst. Appl. 36, 2352

(2009).19. P. Sauseng, B. Griesmayr, R. Freunberger, and W. Klimesch, Neurosci. Biobe-

hav. Rev. 34, 1015 (2010).20. B. Zoefel, R. J. Huster, and C. S. Herrmann, Neuroimage 54, 1427

(2011).21. A. Fink, B. Graif, and A. C. Neubauer, Neuroimage 46, 854 (2009).22. J. Kaminski, A. Brzezicka, M. Gola, and A. Wróbel, Int. J. Psychophysiol. 85,

125 (2012).23. B. Penolazzi, C. Spironelli, C. Vio, and A. Angrilli, Behav. Brain Res. 209, 179

(2010).24. L. S. S. Bialoskorski, J. H. D. Westerink, and E. L. V. D. Broek, Mood

swings: An affective interactive art system, ICST Institute of ComputerScience Social Informatics Telecommunication Engineering 2009 (2009),pp. 181–186.

25. D. Handayani, H. Yaacob, A. W. Abdul Rahman, W. Sediono, and A. Shah,Systematic review of computational modeling of mood and emotion, 5thInternational Conference Influence Communication Technology Muslim World,November (2014), pp. 1–5.


3001


R E S E A R CH A R T I C L E


Vol. 21, 3002–3006, 2015

Making a Successful Agile Team

Beni Suranto

Department of Informatics, Faculty of Industrial Technology, Universitas Islam Indonesia,Yogyakarta 55584, Indonesia

This paper is mainly discuss about Agile software development and strategy to build a successful agile team fordeveloping high quality software products. This paper starts with general information about agile developmentand two most popular methods, Extreme Programming (XP) and Scrum. XP has five core values and twelvebest practices inspired by the Agile Manifesto with the main goal is to organize software engineers so that theyable to produce software products with higher quality within short functionality delivery schedules. Scrum isan agile method which use a software development framework called sprint with a small, self-organizing andempowered team with rapid adaption and complete visibility. Then we will move onto the detail informationabout characteristics of Agile projects which are blurring roles, continuous development activities, and teamaccountability. The next part of this paper discuss about four key factors that have to be concerned by softwareengineers for making a successful Agile team. The key factors are co-located team, engaged customers, self-organizing team, and accountable and empowered team.

Keywords: Agile, XP, Scrum, Software Team, Success Factors.

1. AGILE SOFTWARE DEVELOPMENTAgile software development was officially defined in a form of

“manifesto” created in February 2001 by a group of 17 noted

software process methodologists. They attended a summit meet-

ing in order to propose a better method for developing “right

and working” software and then they formed the Agile Alliance.1

They signed a manifesto contains four fundamental philoso-

phy of Agile software development. This manifesto represents

a new paradigm to effectively change the "too ordered" for-

mal approaches into the "more flexible" one. But it also taking

a stance against the bad characteristics of the ad-hoc method

in software development (e.g., low-quality software products,

undisiplined software engineers).2 The manifesto is shown in

Figure 1.

Agile development method is a response for the customers’

expectation for having high quality software products that meet

their requirements—and soon. It allows the project team to

reduce the cost of change throughout a project by using the fol-

lowing strategies:3

• Produce the first delivery in weeks, to achieve an early win

and rapid feedback

• Invent simple solutions, so there is less to change and making

those changes is easier

• Improve design quality continually, making the next story less

costly to implement

• Test constantly, for earlier, less expensive, defect detection

Nowadays there are many software development methods that

can be called “agile.” Following are some of them:3

• Extreme Programming (XP)

• Scrum

• Dynamic System Development Method (DSDM)

• Feature-Driven Development (FDD)

• Adaptive Software Development (ASD)

• Crystal

• Lean Software Development (LD).

Currently, XP and Scrum are the most two popular Agile meth-

ods. XP which was invented by Kent Beck is a software devel-

opment methodology with collection of values, principles, and

practices. XP shares the values inspired by the Agile Manifesto

with the main goal is to organize software engineers so that they

able to productively develop software products with higher qual-

ity within short functionality delivery schedules.5

XP is based on the following values:6

• Simplicity: In XP, software engineers will only do what is

needed and asked for to produce high quality software products

for reasonable costs. In every iteration, the XP team will focus on

a few requirements and mitigate failures as they happen. There

is one popular principle in XP called "You aren’t gonna need

it" (YAGNI) which states that a software engineers not add new

functionalities until they are required by the customer.

• Communication: Communication is key factor that determine

the agility of the XP team. In XP, software engineers have to

communicate face to face daily to discuss ideas and bring solu-

tions to problems during the project.

• Feedback: In XP, every iteration aims to deliver working soft-

ware product. This can be achieved by demonstrate and test the

3002 Adv. Sci. Lett. Vol. 21, No. 10, 2015 1936-6612/2015/21/3002/005 doi:10.1166/asl.2015.6484


Fig. 1. Agile manifesto.4

software product early and often to get feedback from customers

and the team itself to improve the quality of the software product.

• Respect: In a XP team, every software engineer contributes

value. He/she gives and feels the respect as a valued team mem-

ber. Software engineers and the customers respect each other in

term of their expertise. Also, management respects the responsi-

bility and the authority of the software engineers over their own

work.

• Courage: During the project, software engineers have to tell

the truth about their progress on tasks assigned to them. There

is no excuses for failure in a XP team since it plan to succeed.

But, software engineers in a XP team should not fear anything

because they work together and they will adaptively respond to

changes. We will adapt to changes when ever they happen during

the project.

XP has 12 practices which can be used by software engineers

to enact the above values in a software development process:7

• On-site customer

• User stories

• Metaphor

• Simple design

• Coding standard

• Pair programming

• Collective ownership

• Testing

• Refactoring

• Small releases

• Continous integration

• 40-hours workweek

Figure 2 is a flowchart which describes about how XP’s rules

work together during a software development process.

Scrum is an agile method which use a software development

framework called sprint. A Scrum project usually has a small,

Fig. 2. How extreme programming’s rules work together.8

self-organizing and empowered team with rapid adaption and

complete visibility. There are three main roles in Scrum which

are product owner, Scrum master, and developmet team. Product

owner in Scrum is like on-site customer in XP, he/she is rep-

resenting the interests of customers. His/her main responsibility

is to define and prioritize the requirements of the system being

developed. The scrum master is the “manager” of the develop-

ment team and responsible to ensure that the development team

effectively perform to achieve the goals of the project. The scrum

master also responsible for teaching scrum to software engineers

on the team so they can enact scrum values and practices. He/she

also must be able to resolve any problems during the project.

The development team consist of professional software engineers

who are responsible for implementing the system. The develop-

ment team in a Scrum project has uniques characteristics which

are cross-functional and self-organizing. The development team

does the actual work of delivering the product increment. All

software engineers in the development team should be available

to the project full time.5

In a Scrum project, there is a series of sprint which is a

2–4 weeks iteration to produce the tangible and tested artifacts.

Following are activities for each sprint in Scrum:9

• Product backlog review: this activity is performed by the

development team together with product owner and scrum mas-

ter by reviewing the product backlog. Product backlog is a list of

user stories representing the required functionalities of the system

being developed.

• Product backlog refinement: this activity is used by the devel-

opment team to discuss with product owner in order to change the

order of the user stories listed in product backlog, remove unnec-

essary requirements, add new functionalities, split and merge user

stories, and determine the user stories that will be finished on the

intended sprint.

• Sprint planning: it is a meeting to produce a list of user stories

will be completed in the sprint and a plan for finishing all related

tasks to complete the intended user stories.

• Sprint: during the sprint, software engineers produce the prod-

uct increment as be determined in sprint planning. Product incre-

ment is the most important deliverable in Scrum project. It has

to be high quality, meet the criterias of "done", and accept by the

product owner.

• Daily Scrum: in this daily the meeting, each of software engi-

neers have to explain about what he/she has finished, what he/she

plans for the next day, and what problems he/she has during

his/her work.

• Sprint review: this meeting is held at the end of each sprint.

On this meeting, the development team demonstrates the product

increment they finished on that sprint to the product owner. Then

the product owner will evaluate whether the product increment

is acceptable or not.

• Sprint retrospective: this activity is performed by the develop-

ment team to review their performance during the last sprint. The

development team discuss to identify potential improvements.

The framework which illustrated how Scrum project being per-

formed is shown in Figure 3.

2. CHARACTERISTICS OF AGILE PROJECTSThere are some special characteristics of Agile projects that

software engineers need to know about it. Understanding the

3003


Fig. 3. The scrum framework.10

differences between Agile projects and projects that use other

“traditional” software development methods is a compulsory step

for software engineers before they start their Agile projects.11

The first and the most important characteristic of Agile

projects that distinguish them with non-Agile projects is that

roles are really blur on them.12 Titles or roles (Developer, Tester,

System Analyst, Database Engineer, User Experience Expert, etc)

are not important on Agile projects. People in an Agile project

must pitch in and do whatever it takes for the successfulness of

the project, regardless of their roles. This situation is illustrated

on Figure 4.

This characteristic doesn’t mean that an Agile project employ

people without specific requirements. Of course people joining

an Agile team must have core competencies and also need to

stick to what they are good at. But on an Agile project there are

no “narrowly defined” roles such as programmer, analyst, tester,

and so on and so forth.

The second characteristic of Agile projects is that they perform

all software development stages (i.e., analysis, design, coding,

and testing) as continues activities.13 In every single iteration,

an Agile team performs analysis, design, coding, and testing

activities for a little number on requirements. In the next iter-

ation it performs those activities again for other requirements.

Those activities are never end until the team successfully deliver

the intended system to the users. This situation is illustrated in

Figure 5.

Fig. 4. Roles on agile projects.12

Fig. 5. The difference between activities on traditional projects and agileprojects.12

From Figure 5 shown above we can see that software devel-

opment activities on Agile projects can’t exist in isolation any-

more. So, all software engineers in an Agile team must be able

to working together daily and perform those continues activities

throughout the project.

The third characteristic of Agile project we need to consider

is that quality is a “one team” responsibility.14 Every people

working on an Agile project is a “Quality Assurance Supervi-

sor” an the team, whether they managing the project, eliciting the

requirements, designing the user interface, or writing and testing

the code. Everyone must contribute for the team accountability.

The different situation is happen in traditional projects where the

quality of the requirements is system analyst’s responsibility, the

quality of the code is the programmer’s responsibility, the quality

of the user interface is the user experience’s responsibility, and

so on and so forth. This difference is illustrated on Figure 6.

3. KEY FACTORS FOR SUCCESSFUL

AGILE TEAM3.1. Co-Location

In term of the location of the people envolved in the project, soft-

ware development team can be divided into two types, co-located

team and distributed team. Both of them can run Agile projects.

But, have a co-located team is a better thing on an Agile project.

It can dramatically improve the productivity of the team.13

Effective communication plays an important role in Agile

projects and there is no better strategy to build a communica-

tive team rather than having everyone sit together in a co-located

workplace. On a co-located team, people can get answers for

their questions quickly, problems can be fixed on the spot, and

Fig. 6. The difference between accountability concept in Agile and tradi-tional methods.12

3004


friction can be minimized. Also, intensive interactions among

team members allows trust to be built more quickly.14

Sometimes there is no choice but having a distributed team

to run the project. Following are two useful steps the distributed

teams can do to improve their productivity:13

• At the beginning of the project, team can reserved some bud-

get to bring everyone together even just for 1 or 2 weeks. Team

members must effectively use that time to know each other and

build “chemistry” among team.

• During the project, team can use possible communication

tools (Skype, video conference, social media) so it seem like a

co-located one.

3.2. Engaged Customers

Engaged customers are some stake holders of the intended sys-

tem who show up to system demos, answer questions about the

requirements, give suggestions and feedbacks, and provide the

guidance for the developer team to build a “right and working”

software.13 Basically they can be considered as members of the

developer team.

Having engaged customers that actively participate during the

project is very important. It is impossible for the team to build

compelling and innovative software without having people who

will use that software as part of the process.

All Agile methods fight hard for customers engagement

through practices. For example, Extreme Programming has it’s

on-site customer and Scrum has it’s dedicated role of product

owner.3

3.3. Self-Organizing

To be able to delivering a high quality software, an Agile team

need to performs as a self-organize team. It means that once the

team understands the goal of the project, every software engi-

neers on the team collectively figure out how to achieve that

goal.11

Self-organization is about how software engineers working

together as a team with their unique talents, skills, and passions,

no matter what their roles, so the team can best deliver the project

to its customers.

Self-organization can be considered as an acknowledgement

that the best way to make a successful team is to let the role

fit the person, not making the person fit the role.15 On an Agile

team, it is not a problem when a programmer involved in the

user interface design process or when a software tester involved

in the requirement elicitation process.

Following are some useful tips to get the team to self-

organize:14

• Team recruit people who capable to initiate ideas, have tech-

nical excellence and creativity, and don’t wait for instructions.

• Team must let everyone proposes ideas, creates the plan,

comes up with the estimates, and take ownership of the project.

• Team must worry less about roles and focus on the continuous

production of working and tested software.

Team must trust all members, encourage them, and empower

them to get the project done successfully.

3.4. Accountable and Empowered

A professional Agile team realizes that the customers are really

concern on the quality of the delivered software. So, everyone

Fig. 7. Daily stand up meeting.16

on the team must support the accountability of the results that

the team produces for the customers. Everyone on the team must

understand that the team has responsibility to deliver value for

the customers during the project.

To be able to accountable, the team must be empowered.

Empowering the team can be done by allowing all members

of the team make their own decisions, take their initiatives, do

what their think is right, and act on their own accord.12 It wil

encourage everyone to solve their own problems without wait for

permission from anyone.

One powerful strategy to maintain the accountability of the

team and to empower everyone on the team is get the team to

demo the software. Putting everyone in front of real customers

and having them to demo their works will effectively making the

team more accountable.15 First, everyone on the team will realize

that the real customers (with real problems) are counting on them

to deliver a right and working software. Second, getting the team

to demo the software to the real customers will be very useful to

collect the feedback that needed to improve the software. Daily

stand up meeting as shown in Figure 7 is an example of activities

that support the accountability of an agile team. Daily stand up

meeting is a 15 minutes meeting which is held by the team at

the same time everyday during the project. During this meeting,

each team members explain three things which are the progress

of his/her tasks, any problem that comes his/her way, and what

he/she will do today.

4. CONCLUSIONSAgile software development becoming a powerful method to

developing high quality software that meet the customers’

requirements and effectively deal with the problems of rapid

change during the project life-cycle.

The project teams which want to use Agile in their project

must first clearly understand about the characteristics of Agile

projects that significantly different from other “traditional” soft-

ware development methods. Some important characteristics of

Agile projects are blurring roles, continuous development activi-

ties, and team accountability.

Some key factors that can be very useful for improving the

team’s productivity and helping the team to successfully achieve

the goals of the project are co-located team, engaged customers,

self-organizing team, and accountable and empowered team.

3005


References and Notes1. A.Alliance,What isAgileSoftwareDevelopment?,AgileAlliance, viewed21Jan-

uary 2015,<http://www.agilealliance.org/the-alliance/what-is-agile/> (2012).2. J. Appelo, Management 3.0: Leading Agile Developers, Developing Agile

Leaders, Pearson Education (2011).3. A. Cockburn, Agile Software Development: The Cooperative Game, Person

Education Inc., Boston (2006).4. A. Manifesto, Manifesto for Agile Software Development, Agile Manifesto,

viewed 21 January 2015, <http://agilemanifesto.org/> (2001).5. D. Leffingwell, Scaling Software Agility: Best Practices for Large Enterprises,

Pearson Education (2007).6. D. Wells, The Values of Extreme Programming, Don Wells, viewed 4 February

2015, <http://www.extremeprogramming.org-/values.html> (2009).7. XProgramming, What is Extreme Programming?, XProgramming, viewed

4 February 2015, <http://xprogramming.com/what-is-extreme-programming/>(2011).

8. D. Wells, Extreme Programming Project, Don Wells, viewed 3 February 2015,<http://www.extremeprogramming.org/map/project-.html> (2009).

9. S. Alliance, Core Scrum, Scrum Alliance, viewed 6 February 2015,<https://www.scrumalliance.org/why-scrum/core-scrum-values-roles>(2014).

10. K. Rubin, Scrum Framework, Agile Atlas, viewed 3 February 2015, <http://agileatlas.org/articles/item/scrum-framework> (2012).

11. T. Dingsøyr, The Journal of Systems and Software 85, 1213,DOI: 10.1016/j.jss.2012.02.033.

12. J. Rasmusson, The Agile Samurai: How Agile Masters Deliver Great SoftwarePragmatic Bookshelf (2010).

13. E. Derby and D. Larsen, Agile Retrospective: Making Good Team Greats, ThePragmatic Bookshelf, Texas (2008).

14. M. Holcombe, Running an Agile Software Development Project, Hogn Wileyand Sons, Inc., New Jersey (2008).

15. J. Shore and S. Warden, The Art of Agile Development, O’reilly Media Inc.,California (2008).

16. A. Miller, How Microsoft’s p&p Teams do Daily Standup Meetings,viewed 2 January 2015,<http://www.ademiller.com/blogs-/tech/2008/07/daily-standup-meetings/> (2008).

Received: 9 November 2014. Accepted: 27 December 2014.

3006




Vol. 21, 3007–3009, 2015

Optimized Performance Result for 2.4 GHz and

2.45 GHz Circularly Polarized Microstrip Antenna

Rudy Yuwono1�∗, Ronanobelta Syakura1�∗, Erni Yudaningtyas1, Endah B. Purnomowati1, and Aisah2�∗

1Department of Electrical Engineering, University of Brawijaya, Malang, 65145, Jawa Timur, Indonesia2Department of Electrical Engineering, Malang, 65145, Jawa Timur, Indonesia

A deployment of the RF devices especially for antenna should be done continuously to obtain the better perfor-mance than the existing. The better performance means that the antenna has the matching impedance closedto perfect condition, higher gain and also better alignment which is related to antenna polarization. To get thebetter performance, the antenna should employ circular polarization which has advantage in alignment. Theoptimization of the antenna should be done to improve the performance. In this paper, the optimized antennacan covers between 2.4 GHz–2.45 GHz of Frequency with the increased gain and circular polarization whichgenerally have better performance compared than existed antenna.

Keywords: Circularly Polarized, Microstrip Antenna, 2.4 GHz, Optimization.

1. INTRODUCTIONA deployment of the RF devices especially for antenna should

be done continuously to obtain the better performance than the

existing. The better performance means that the antenna has the

matching impedance closed to perfect condition, higher gain and

also better alignment which is related to antenna polarization.

The antenna type should be fabricated in low cost and easily.

Thus, the microstrip antenna can be used because it can meet

both criteria.

The antenna should be applied into wireless technology

directly. There is a frequency spectrum band which has free

license status between 2.4–2.5 GHz. The frequencies of 2.4 GHz

and 2.45 GHz are applied for ISM band application such as Wi-Fi

and RFID detection system which use low power.

The alignment of the antenna also takes effect when the infor-

mation transmitted. If the antenna employs linear polarization for

data transmission, it will lead the loss of data because of the mul-

tipath fading.4 To solve this problem the antenna should employ

circular polarization which has advantage in alignment.

2. THE ANTENNA DESIGNThe design of the antenna was using circular shape as Ref. [8]

and modified by adding the slot and patch reposition to

obtain the antenna has vector phase 900 which results circular

polarization.1�4�5 The antenna model is shown at Figure 1.

∗Authors to whom correspondence should be addressed.

The existed antenna as explained at Ref. [8] has the dimension

of R = 20 mm, x = 9 mm and length of m = 4�28 mm and

n = 18 mm. The aim of the antenna design modification is to

improve the matching impedance which takes effect to bandwidth

improvement.

There are two modified antennas are designed. The antenna is

modified by adding the T -shape slot at ground plane, resizing

the circular patch-x and resizing the length of n. The modified

antenna result is shown at Figures 2 and 3.

From Figure 2, at groundplane of the modified antenna, there

is T -shape slot addition to match the impedance. The dimension

of T -shape slot length of a= 14 mm, b = 5 mm and c = 4 mm.

Meanwhile at Figure 3, the patch of the antenna has 2 models;

the one is resized n part (on Fig. 3(a)) and the other by shifting

the x part which has circle shape (on Fig. 3(b)).

Both antennas use the T -shape at the groundplane. The length

of n part is resized to 17 mm and 16.5 mm on each antenna

model. And the x-part shifting only applied at the antenna which

has modification of the n length at 18 mm to ensure the vector

still at around 900 of phase.

3. OPTIMIZATION RESULT

The optimization result employs three antennas, the existed

antenna (et) which obtains as Ref. [8], the modified antenna,

which has resized-n part at 17 mm of length, un shifted x-part

(mod 1), and the one which has resized-n resized part at 16.5

mm, shifted x-part 0.5 mm (mod 2). The result only covers until



Front Rear

Fig. 1. The existed antenna model.8

the frequency of 2.5 GHz as the focus on ISM band. The opti-

mization results are shown at Figures 4–6.

The result of the S11 from Figure 4 explains that the mod-

ified (mod 1 and mod 2) antennas have the advantage than

the existed (et) antenna. For two modified antennas have bet-

ter matching impedance, which indicate from the S11 result least

than −9.54 dB as the tolerance reference of mismatch,7 however

the S11 result only conducted between 2.35 GHz to 2.45 GHz.

From Figure 5 the gain of the modified antennas is increased.

Those are contradictive to existed antenna which has decreasing

trend.

The results from the Figure 6 obtain that the existed antenna

can covers the axial ratio least than 3 dB at each frequency which

indicates that the antenna has circular polarization. It indicates

Fig. 2. Rear view of the modified antenna.

(a) (b)

Fig. 3. Modified antenna models.

Fig. 4. S11 result of the each antenna.

Fig. 5. Gain result.

Fig. 6. The axial ratio result.

different result at modified antennas, which covers 2.35–2.4 GHz

of frequency (mod 1) and covers 2.45 GHz of frequency (mod 2).

4. CONCLUSIONThe antenna improvement results that the modified antenna can

cover 2.4 GHz to 2.45 GHz of frequency. Each antenna has dif-

ferent specification. Both of the antennas have increased gain.

comparing with the existed antenna the two modified antennas

can perform better, however the circular polarization only cover

2.35–2.4 GHz of frequency for first antenna model and 2.45 GHz

of frequency for second antenna model.

References and Notes1. C. A. Balanis, Antenna Theory Analysis and Design, Wiley, USA (2005).2. P. Bhartia, I. Bahl, R. Garg, and Ittipiboon, Microstrip Antenna Design

Handbook, Artech House, USA (2001).

3008


3. M. Haneishi and Y. Suzuki, Circular polarization and bandwidth, IEE Electro-magnetic Series 28-Handbook of Microstrip Antennas, edited by J. R. Jamesand P. S Hall, Peter Peregrinus, London (1989).

4. R. Joseph, Studies on Circularly Polarized Broadband Slot Antennas, Ph.D.Thesis, Kumamoto University, http://hdl.handle.net/2298/22943, Japan (2011).

5. R. Joseph and T. Fukusako, Progress in Electromagnetics Research C 26, 205(2012).

6. G. Kumar and K. P. Ray, Broadband Microstrip Antennas, Artech House, USA(2003).

7. S. Makarov, Antenna and EM Modelling with Matlab, Wiley, USA (2002).8. R. Yuwono, R. Syakura, and D. F. Kurniawan, Design of the circularly polar-

ized microstrip antenna as RFID tag for 2.4 GHz of frequency, InternationalConference on Advances Technology in Telecommunication Broadcasting andSatellite 2014, Bali-Indonesia (2014).

Received: 12 September 2014. Accepted: 12 October 2014.

3009




Vol. 21, 3020–3024, 2015

A New Technique for Protecting Server Against

MAC Spoofing via Software Attestation

Kamarularifin Abd Jalil1, Nor Shahniza Kamal Bashah1�∗, and Mohd Hariz Naim @ Mohayat2

1Faculty of Computer and Mathematical Sciences, Universiti Teknologi MARA, 40450 Shah Alam, Selangor, Malaysia2Faculty of Information and Communication Technology, Universiti Teknikal Malaysia Melaka, Hang Tuah Jaya,

76100 Durian Tunggal, Melaka, Malaysia

A server is always vulnerable to network attack especially MAC spoofing activity. Current existing MAC spoofingdetection algorithms have several flaws that make protecting the server hard to implement. Thus, a new methodfor detecting MAC spoofing attack is proposed in this research where it is able to detect any spoofing activityand also pinpoint the attacker with minimal false positive.

Keywords: Spoofing, Algorithm, Attack, Genuine, Attestation.

1. INTRODUCTIONMAC address is a unique identification for any network interface

card (NIC) which means that one MAC address should belong

to only one NIC. However, this MAC address can be changed by

certain techniques especially when using software, making two

or more NIC having same MAC address. This is called MAC

Spoofing where an attacker would exploit to execute denial of

service by broadcasting false messages causing interruption to

server service. Denial of Service (DoS) attacks is commonly hap-

pen to a server where it would interrupt the service making it

unavailable to be accesses. File and email servers are the most

targeted victims and one of the attack technique to perform DoS

is by MAC spoofing. It is well known that MAC address spoofing

could lead to many network attacks and some current techniques

to detect MAC address have their own weakness. For example,

as described in Refs. [1, 2], to detect spoofing from received

signal strength, the genuine station must be stationary. However

the result might not be accurate if the station is on the move or

mobile. Furthermore, as mentioned by Ref. [3], spoofing attacker

might avoid being detected by stopping to reply to any frames.

This could be done by modifying the settings of the NIC mon-

itoring stations. In general, determining which host is genuine

and which one is the attacker could not very accurate. As a result,

network attacks could still happen if MAC address spoofing is

not detected and identified precisely.

This paper will proposed a new technique for detecting MAC

address spoofing. It starts with the study and evaluation of other

techniques for detecting MAC spoofing. Next, the main part


of the paper explains thoroughly on the proposed method for

detecting the MAC spoofing attack. The experimental results are

presented in the later section to illustrate the effectiveness of

the new proposed method. Future works are summarized in the

conclusion.

2. RELATED WORKSIn order to avoid re-inventing the wheel all the relevant existing

techniques for detecting MAC spoofing will now be successively

examined thoroughly.

2.1. Sequence Number Analysis

According to Ref. [3], the sequence number analysis is based on

MAC header sequence number fields that reside in Data Frames.

It is a unique number of sequences which will increments each

time a frame is sent out by a node. This sequence number is

mainly used to combine back the fragments of MAC frame. This

uniquely generated number is the key for detecting MAC spoof-

ing even though an attacker is capable of forging the sequence

number. As stated by Ref. [3], to perform spoofing attack, an

attacker needs to imitate the source address and also the corre-

sponding sequence number. In order to detect spoofing, there will

be monitor node that perform checking to the sequence num-

ber. From the sequence number analysis’s theory, the number

gap of inter-frame sequence should be equal to one; and spoof-

ing is detected whenever the sequence number is not equal to

one. However, in real situations the number gap sometimes may

not be equal to one as frames may be retransmitted due to lost

during transmission and may lead to high tendency of false pos-

itives. To overcome this problem, this technique will perform



some checking for Duplicated Sequence Number via an algo-

rithm of a pseudo code. In addition to the pseudo code checking,

the sequence number gaps between frames are now between 0

and 4 to overcome the problem of duplicated or retransmitted

frame. The monitor node will also send ARP request to the trans-

mitting node and at the same time stores the current sequence

number gap value. At this stage, the node will be labeled as in

verification state where the monitor node continuously checking

for the next sequence number is same with the current sequence

number. In fact, the sequence number analysis majorly just per-

forming detection whether MAC spoofing attack exists or not.

It does not help to identify and differentiate between a legiti-

mate node and spoof attacker. The detection algorithm that relies

on the gap number between frames transmitted which results to

the tendency of false positive and false negative. Also, bear in

mind that certain open-source drivers and modified firmware of a

device do allow sequence number manipulation.2 There are also

MAC-layers frames that do not contain sequence number and

thus making the detection less accurate. Another crucial limita-

tion of the sequence number analysis is that it also relies on ARP

request and response which requires the nodes to be within the

same network segment. If the attacker does not presence in the

same network segment as the legitimate node, the detection node

is not capable of detecting MAC spoofing attack.

2.2. MAC Spoofing Detection Using PLCP Header

This technique use the MAC frames to distinguish between gen-

uine node and attacker. The MAC frames that are send by nodes

like Acknowledge frames Clear to Send Frame and authentication

frames shall be used as detection mechanism. The term PLCP

header here refers to all the frames mentioned before and this

PLCP header are much harder to forge as claimed by Ref. [1]

because of adaptive algorithm and environments of the NIC. The

PLCP header consists of two main parameters which is data rate

and modulation types. Although the attacker can avoid being

detected by stopping to replay the ACK frames, it would even-

tually limit the capability of performing attack. To experiment

with the mechanism, Chumchu et al. set up the test bed where

the monitoring station sent out Data Frame to the specific gen-

uine node. Note that the attacker had already changed the MAC

address to be similar with the node station. Upon receiving the

data frame, both of the nodes will reply ACK frames to the

monitoring station. From here, it is observed that the radio tab

headers are quite different and thus it means that one of node

must have spoofed the MAC address. Chumchu et al. also tested

out the performance of using the technique and the result shows

that spoofing detection could not be 100 percent accurate. This is

due to the sequence number of both legitimate node and attacker

which may be same at certain time. Despite the results shown

above, the technique is only meant for detecting spoofing attack

activity but to actually pin point which one is legitimate node

and which one is attacker node is seems to be impossible. In

addition, if the attacker stops from replying the ACK frame,5 the

spoofing attack still cannot be detected.

3. THE PROPOSED ALGORITHMTo test out the new method effectiveness in detecting spoofing,

an experiment will be conducted based on the attack model pro-

posed by Sheng2 as depicted in Figure 1.

A legitimate node scanand connect to thedesignated SSID

The legitimate nodeconnect to the detector

via HTTP request

Detector save nodefingerprint based on User

Agent Information andMAC address

Node fingerprint alsoplanted into client side

Attacker changed MACaddress and connect

to the detector

Duplicated MAC will bedetected and request for

fingerprint information

Attacker will failauthentication as

fingerprint informationis wrong

Fig. 1. General flow of the experimentation.

First, a legitimate node will make connection to the detector

for initial connection. Detector will then save the node fingerprint

based on User Agent Information (UAI) and MAC address. The

information will also be planted into the node. The information

that is planted into the node is like a key for future identifica-

tion and authentication that will be requested by detector. Next,

the attacker will change the MAC address to be the same with

the legitimate node and then will try to make connection to the

detector for authentication. Detector then detected that a duplica-

tion is exist in the network and start request for node fingerprint.

Since the attacker does not have the correct key, the attacker will

be labeled as spoofer and can be then blocked from accessing

the server.

The new MAC spoofing detection method would be derived

from the concept of three existing methods. Instead of modify-

ing and altering the device firmware or develop a customized

hardware, the new proposed method will be using existing fea-

tures and available information from the node to create a key-like

fingerprint. The proposed detection method will also be using

ARP information to collect list of MAC address that exist in the

network segment. The ARP information will be run on detector

server and extract and process the information to check for dupli-

cated MAC address existence in the network. Firewall or other

intrusion detection mechanisms are also not needed in the pro-

posed detection method. In general, the basic requirements that

will be needed for the technique are as follows:

i. ARP information for detecting MAC spoofing activity

ii. User/node information for uniquely identify the user/node

iii. An access point or wireless switch/router

iv. A detection server

v. 2 Nodes/Machines to simulate a legitimate and attacker.

3021


4. EXPERIMENT AND RESULTSIn this chapter, the results of MAC address spoofing detection

are discussed in details. First of all, the MAC addresses need to

be captured by running ARP command. The User Agent (UA)

is used to store visiting node information. The UA is consists of

some information which is used for identification of a visiting

node. Both the UA information and MAC address of the specific

node are saved into the database. Before running the experiments,

some discussion on the UAI is elaborated in details in the next

subsection. The steps to obtain and process the information are

also mention in the remaining subsection.

4.1. User Agent Information (UAI)

Since the proposed method is derived mainly from the

Transceiver Fingerprint, where a node will be assigned to a

unique key for identification, some information from the node

itself is required for making the unique key. From some research

that has been done, it is discovered that a node may have infor-

mation each time it access a server called UA. The UAI is the

information that consists of browser information such as the ver-

sion, type and operating system. This information together with

MAC address can be used for creating a fingerprint like key

which become unique each time the node is connecting via HTTP

request. The example of UAI that is requested is as follow:

Mozilla/5�0 �X11; Ubuntu; Linux i686; rv:12�0) Gecko/20100101 Firefox/12�0,00-1e-65-c8-2c-6d

Here, the information means that the node is using Mozilla

browser to perform HTTP request and the node operating sys-

tem is running on Ubuntu Linux with processor intel686. This

information if combine with MAC address would create a unique

key to identify a node. Then the information will be stored into

the database and also at the node side that is embedded into the

node’s browser cookies. Later on the information can be retrieved

by detector for authentication purposes.

4.2. ARP Request and Response

Address Resolution Protocol (ARP) is a communication proce-

dure to resolve network layer address into link layer or in the

other words; it is used for translating IP address to MAC address

so that devices within the same network segment can identi-

fied their neighboring nodes. Since the proposed method is using

MAC address information for detection purposes, ARP request

and respond is used. In windows operating system, ARP request

can be executed by invoking ARP—a “IP address of Node.” For

example by running the ARP—a 192.168.1.33 on Windows ter-

minal, it will receive ARP response of 00-1E-65-C8-2C-6C and

this is the MAC address of the specific node. Meanwhile run-

ning ARP—a without specifying the node’s IP address will return

ARP response that have list of MAC addresses as presented in

Table I.

Table I. ARP list of MAC address.

IP Address MAC address

192.168.1.1 B8-Af-67-F2-D3-03192.168.1.22 00-0F-FE-7B-F4-D5192.168.1.5 00-26-9e-A5-4a-A5192.168.1.33 00-1E-65-C8-2C-6C

4.3. Proposed Method Experimentation Steps

In order to literally scan for spoofing activity in a network, the

proposed detection method would need to base on the connection

flow as depicted in Figure 2. At first, a genuine initiates HTTP

request connection to the detector/server. Upon connected, the

detector then triggers the ARP request to the connected client to

get the MAC address information. At the same time, the detector

also invoke for UAI that is available via node’s browser. After

has successfully get the MAC address and UAI from the node,

the information is combined together to create a key-like finger-

print that is used later on during authentication. The unique key

is then saved into database and also saves back to the node’s

browser cookies. In the next phase, the attacker will change its

node MAC address to be similar with the legitimate node and

then tries to communicate with the detector/server via HTTP con-

nection. Then, the detector/server trigger again ARP request and

soon discover that duplicated MAC address exist within the net-

work. Now that the duplicated MAC address has been detected,

the detector/server then try to request the key fingerprint that has

been embedded into browser cookie previously. At this stage,

the attacker fails to provide the key fingerprint as similar with

the one in the database and detector can pinpoint that node is

actually a spoofer.

We begin by specifying the notation that will be used.

By running the ARP command, a list of MAC address can

be obtained. The MAC address will then be saved into database

for matching up and authentication. In this research, ARP—a

and ARP—a “Node_ IP_ Address” are used. The ARP—a will

return a list of MAC address and IP address within the same

segment. Meanwhile ARP—a “Node_ IP_ Address” will only

return one MAC address of the specific IP address. For example

running ARP—a 192.168.1.33 will only return one MAC address

of 00-1E-65-C8-2C-6C.

4.4. Setting up Network Test bed

Before MAC address can be captured, the network is setup via

a wireless router or switch. In this test scenario, to imitate a

spoofing attack both node B and C are virtual machine running

Ubuntu 12.04. First, node A is start up and connects to the Access

Point (AP). The IP address is assigned as 192.168.1.34 by the AP.

After has successfully connected, only then the Node C (attacker)

is start up and tried to connect to the AP. This is to mimic that

Fig. 2. Logical flow in detecting spoofing.

3022


Fig. 3. Network configuration of node B.

the attacker has already observed and know the MAC address

of node B and thus change his/her node MAC address to be the

same with Node B. The attacker has auto-assigned IP by DHCP

or can assigned IP address manually in case of getting conflict

IP address. In this case, the IP address is manually change by

attacker to 192.168.1.35.

4.5. Changing MAC address of Virtual Machine (VM)

To imitate a spoofing attack, the MAC address needs to be

changed so that there is more than one similar MAC address.

Using the Oracle VM Virtual box, the MAC address of VM can

be modified before starting up. Both of the machines are actually

running from Windows platform at Node A. Then, each of the

VMs network adapter connections are set to Bridge Adapter (BA)

where it is virtually making both machine attached to the AP. In

this scenario, the virtual machine’s MAC address is changed to

00-1E-65-C8-2C-6D similar to Node B.

Figure 3 shows the IP address and MAC address of Node B.

The IP address is 192.168.1.34 while the MAC address is

00-1E-65-C8-2C-6D.

Figure 4 shows the network configuration of the attacker

Node C. Take note that the MAC address has been changed to

be similar with Node B at Figure 3.

4.6. Running ARP Command

Now that the network configurations are all set up and MAC

address had duplicated, the detector Node A should be able to

detect its neighboring nodes. Figure 5 presents the returned MAC

addresses and from here the duplicated spoofed MAC address is

also detected. In order to extract the information so that it can be

Fig. 4. Network configuration of node C.

Fig. 5. List of MAC after running ARP command.

differentiated between legitimate and attacker, some processing is

done at the UAI. The information is retrieved each time the node

make connection to the detector via HTTP request. The informa-

tion are then stored into database for information matching.

4.7. Extracting Information and

Authentication Algorithm

The information retrieved from the ARP command above is then

processed through an algorithm as illustrated in Figure 6. Upon

connecting to the detector or server, the node will be perform-

ing some checking process for authentication purpose. First, the

detector at the server will check whether the MAC address is

registered or not. Initially, if the node has not yet makes any con-

nection, the node MAC address and UAI is registered into the

database. The UAI is also embedded within the node browser.

On the other hand, if the MAC is already registered, the detector

then continues to make inspection whether there is any dupli-

cated MAC detected. If any duplicated MAC is detected, then

the detector request for UAI that was embedded during the ini-

tial connection. The information then will be compared with the

Node Connectto Detector

User Agentinformation is

planted within node

Register MACaddress

MAC address

Is MACregistered?

Isduplicated MAC

Detected?

Detector requestUser Agentinformation

Is informationmatch withdatabase?

Alert/Block Nodefrom Accessing

Server

No

Yes

Yes

No

Yes

Node Informationand Time Connected

Fig. 6. Algorithm overview of spoofing detection.

3023


information from the database. If the node fails to supply with

the correct UAI, the node is categorized as a spoofer and then

can be straight away blocked from accessing the server. The node

information such as the IP address and time of connection is also

stored in the database for logging purposes.

Once the system detected that Node C is using similar MAC

address as Node B, a warning message will appear on the screen.

It indicates that the detector senses the duplicated MAC address

value and try request for UAI. Now that the Node C did not have

the same UAI as stored in the database, thus it will get warn-

ing message and the service can be blocked from accessing the

server. The information will also be recorded into the database

for logging purposes.

5. CONCLUSIONSIt can be concluded that detecting MAC spoofing and determin-

ing which one is the attacker is possible via software attestation.

The proposed technique is able to capture MAC address and pro-

cess the information for authenticating and validating connected

nodes. Thus by implementing this technique, any server that pro-

vides services within local network can be protected from MAC

spoofing attack. Network administrator can also be notified in

case of spoofing detected and able to determine which IP address

is performing spoofing attack. In the forthcoming, the proposed

technique of detecting MAC spoofing will be enhance to cope

with its limitation and constraints especially to cater the problem

of different browser UA at client side. It is also hope that the

technique can be used in different network segments particularly

over the Internet for making automatic authentication via MAC

address. Lastly, the problem with MAC spoofing can be solved

or reduce with this new implementation technique.

Acknowledgments: The authors would like to thank Uni-

versiti Teknologi MARA and Malaysian Ministry of Higher Edu-

cation for funding this research under the Fundamental Research

Grant Scheme.

References and Notes1. P. Chumchu, T. Saelim, and C. Srikauy, A new MAC address spoofing detecion

algorithm using PLCP header, IEEE ICOIN (2011).2. Y. Sheng, K. Tan, G. Chen, D. Kotx, and A. Campbell, Detecting 802.11 MAC

layer spoofing using received signal strength, INFOCOM (2008).3. F. Guo and T. Chiueh, Sequence number-based MAC address spoof detection,

8th International Symposium (2006).4. R. A. Redner and H. F. Walker, SIAM Review 26, 195 (1984).5. J. Bellardo and S. Savage, 802.11 denial-of-service attacks: Real vulnerabili-

ties and practical solutions, Proceedings of the USENIX Security Symposium,Washington, D.C., August (2003), pp. 15–28.

Received: 7 November 2014. Accepted: 27 December 2014.

3024




Vol. 21, 3025–3029, 2015

Precursor Emotion of Driver by Using

Electroencephalogram (EEG) Signals

Norzaliza Md Nor1�2�∗ and Abdul Wahab Bar1

1Department of Computer Science, Kulliyyah of Information and Communication Technology, IIUM2Faculty of Medical and Bioscience, University Technology Malaysia, UTM

Driver behaviour is indeed reckoned to be one of the highest factors affecting fatal accidents. However, majorityof the cases can be avoided if the driver can remain focus and make a correct decision in controlling thevehicle while driving. Decision-making ability of the driver is impeded due to driver behaviour which may involveprecursor emotion of the driver that could lead to fatal accident. Thus, understanding and analyzing the driverbehaviour and the resulting emotion can help prevent accident and reducing accident fatality rate. In this paper,the understanding of precursor emotion of driver is studied in details. This correlation between precursor emotionand their respective emotion can be analysed based on the 2-D Affective Space Model (ASM) using four basicemotions (happy, calm, fear and sad) as stimuli. In this case, the Electroencephalogram (EEG) device is usedto extract brain waves signal while the driver is driving the simulator. The EEG signals are captured through thescalp of the driver and features is extracted using Mel Frequency Cepstral Coefficient (MFCC). Neural networkclassifier of Multilayer Perceptron (MLP) is used to classify the valence and arousal axes for the ASM. Analysisof the precursor emotion for driver shows an interesting finding that complements the discrete classification.In addition, the analysis also indicates how precursor emotion can affect driver behaviour. Consequently, theunderstanding of pre-cursor emotion and its relationship towards driver behaviour could help the driver to controlhis/her emotions while driving which can prevent to fatal accident.

Keywords: Precursor Emotion, Driver, MFCC, Neural Network.

1. INTRODUCTIONAccidents happen for many reasons. Some of the features are;

road condition and the environment, the human factors and the

vehicle factors which can affect the driver’s action while driv-

ing. Nevertheless, human factor occupies the largest portion of

the total number of traffic accidents and can be categorized in

their impacts; namely: deaths, heavily injured and slightly injured

which remain above 90%.1 Driving is a risky decision making

process and puts the driver under stressful circumstances. Hence,

each decision which made by the driver will affect his or her driv-

ing manoeuvres. Therefore, the precursor emotion of the driver

is important in this research study.

Driver’s behaviour change does not depend solely on the

driver’s present emotional state but also on the environment and

the driver’s prior emotion. According to Clark,2 emotions are

central to human motivation: they are both the precursors to, and

an end result of, many undertakings. Therefore, precursor emo-

tion of the driver may influence the driver’s emotion as it has

been derived before he/she starts driving. Thus, each driver may


have their own precursor emotion that will determine the exact

emotion they have. Consequently,2 defines emotions as highly

personal in nature: one man’s meat is another man’s poison which

means what is pleasant to one person may be unpleasant to

another (American Heritage Dictionary, 2005).

The elementary architecture of the driver cognitive system has

been illustrated in Figure 13�4 which depicts the driver mental

representation as the correlation between working memory (WM)

and long term memory (LTM) and its environment.3 This archi-

tures has examplify the relationship between emotion and driver

behaviour to achieve the driver’s goal based on the active knowl-

edge in WM through mental representation and reasoning pro-

cess. Besides, LTM stored the driving knowledge permanently

that has been received from learning process which will influ-

ence the driver’s action as well. Moreover, the precursor emotion,

cognitive resources or pre-emotional hijack might be influence

the reasoning process which includes decision making, anticipa-

tion and action planning to activate the knowledge. However, the

information from the environment will also complemented with

the driver’s action in controlling the vehicle. In conclusion, we

can assume that any disturbance to the driver can be potentially

lead to driver’s action.



Fig. 1. Elementary architecture of the driver cognitive system.4

Emotional events usually obtain an excellent position in

memory. There have been many researches conceptualizing emo-

tion along two dimensions of valence, which depicts the plea-

sure or sadness and arousal depicts of calmness or excitation.5�6

Yet tremendous work has been accomplished based on neural

responses under memory formation of emotional events, how-

ever the study on precursor emotion effects by using Electroen-

cephalogram (EEG) is scarce. In this research, a novel approach

has been proposed to analyze and understand the precursor emo-

tion of the driver based on the affective space model (ASM)

that enables emotions to be viewed in two different axes of the

valence (V) and arousal (A) as the essential basis function.13

The valence arousal approach (VAA) can also help to analyze

and visualize the pre-emotion of the driver in driving simulation

task. It will reveal the correlations between precursor emotion

and the pre-emotion towards driver’s behaviour especially in pre

and post-accident which will be studied in greater detail. Thus, in

this research study, the understanding of precursor emotion effect

towards driver behaviour through affective space model will be

analyzed to establish driver profiling. In addition, the study of

the driver emotional states with respect to the change of different

driving environment can provide a better understanding in ana-

lyzing the driver behaviour; hence the effect on precursor emo-

tion towards early detection of highly emotional agitated driver

will be identified to prevent an accident.

2. EMOTION AND EEGElectroencephalogram (EEG) is a tool used to record the electri-

cal activity of the brain.10 The electrical activity is produced by

the brain cells (neurons) and neural circuits. EEG will be applied

in this research to estimate the neural activity.7 For more than

two decades, the individual’s emotional state is learned by using

EEG. Thus, EEG has been adopted to capture neural activity on

a millisecond scale from the entire cortical surface.8

3. PRECURSOR EMOTIONPrecursor is defined as something that existed before and incor-

porated into something that come later (American Heritage Dic-

tionary, 2005). That will be in the general context of precursor

itself but it has quite similar meaning to precursor emotion.

According to Sroufe,9 precursor emotion is something that can

form the basis of interpersonal regulation to start. In other hand,

precursor emotion is the reactions appear before it is being pro-

ceeds to the real emotion. Eyes closed data will be used to

identify the precursor emotion as it represents the initial emo-

tional state of the driver. In this research study, the correlation

between precursor emotion and driver behaviour are analyzed to

generalize the driver behaviour profiling.

4. EXPERIMENTAL DESIGNTo properly place the electrodes on the scalp of the participants,

Nuprep electro-gel was used to clean the scalp surface, this it

to ensure the sensors are connected well to the scalp with the

impedance of the electrodes is minimized. In order to make the

sensors stick well onto the scalp, the Ten20TM conductive gel

was used. The viscosity paste is needed to ensure that the gel

will not flow easily and yet it can be easily remove. Besides,

Ten20TM conductive also helped to further reduce the impedance

on the sensors. While setting up the electrode, the driver is famil-

iarize with the driving simulator. Data collection will be divided

into two parts; basic emotions and driving task. The experimental

design flow is shown in Figure 2.

The first task is the driver need to perform eyes open and eyes

close for one minute. During eyes open task, the driver will be

looking at the blank white screen to represent emotion calm. In

contrast, eyes closed data will be used to analyze the precursor

emotion effect on driver behaviour. In addition, eyes open and

eyes closed data is also use as an EEG data initialization. The

three movie clip of basic emotion will then be shown for one and

half minute per emotion representing emotion happy, fear and

sad. After each movie clip, the driver is required to fill up the

SAM (Self-Assessment Manikin) as shown in Appendix V for

emotional evaluation. Finally, the driver is requested to accom-

plish three different driving tasks and the recorded brainwaves

are then stored onto a hard disk for further analysis.

In this research, 3 types of different driving environment are

set up. These environments represent diverse level of difficulties.

Driving Task

Emotion movie clip

10 minutes per task

1½ minutes each

Eyes close Eyes open (Calm)

1½ minutes each → SAM→ Rest

Electrode’s placement

Brief from the experimenter

Happy Sad Fear

Fig. 2. Experimental research design.

3026


The first task will be driving under normal road traffic in order to

familiarize the driver with the driving stimulator while listening

to the annoying sounds such as dogs barking, ambulance sounds

and baby’s crying. This is designed to identify the fluctuation

of the driver emotion when the driver is distracted. Whereas, in

second task requires the driver is required to drive in a medium

congested traffic and the driver needs to answer several questions

regarding his/her driving experiences. This is to distract the driver

attention by putting him/her under stress condition. Finally, the

driver needs to drive in a heavily congested traffic.

4.1. Stimuli

In order to get emotional responses, the emotion’s movie clip

is used in this research study.13 Three basic emotions will be

displayed to the driver. They include happy, fear and sad emo-

tion by using the International Affective Picture (IAPS). Bernard

Bouchard’s synthesized musical clips and Gross and Levenson’s

movie clips which can be used to elicit emotional responses.11�13

Then, the driver need to drive in three conditions of driving;

task 1—easy driving, which includes the noisy sound to disturb

the driver, task 2—bulked driving, in this part the experimenter

will interviewed the driver in order to see how the driver answer

the question, and finally task 3—heavy driving, congested traf-

fic is purposely designed to see the driver driving skill and their

behavior while having a stress condition.13 The driving task has

been accomplished by using driving simulator 2009.

4.2. Participants

10 healthy drivers (6 female and 4 male) have been recruited

from the final year students of International Islamic University

of Malaysia (IIUM). All the drivers have at least 3 years driv-

ing experience and a valid driving license. The drivers must be

within 20 to 26 years old, since the target group is the young

and healthy driver.

100 C3

10 C3

10 C4

10 T3

10 T4

Decimate[5000features]

MFCC(110

instances)

KDE(128

instances)

Classifier[Trainingnetworkbased onEmotionData persubject

(randomize)]

[15000features]

V

A

100 C4

100 T3

100 T4

Networkgenerator for

differentexperiment

Feature ExtractionRaw Data Pre-processingClassification

Result

Fig. 3. Training data block diagram for generating network classifier.

4.3. Experimental Setup for Training and

Testing Data

Emotion data is trained to establish the network that will be used

in every experimental design. Each experimental design is differ-

entiated based on the size of data and the method used to get the

result. In Figure 3, four steps are needed to establish the network

in training the emotion data for valence and arousal.

Results can be generated using the combination of these four

steps. However, the size of data for each experiment is not being

discussed in details. As shown by Figure 3, the 15000 raw data

that we get by multiplying sampling rate (250) with 60 seconds

of emotion data is first pre-processed by low pass filtering and

down sample from 250 samples per second to 83 samples per

second. This is to remove the entire high frequency noise arti-

fact, since only alpha and beta band are of interest in the emo-

tional analysis.12 The Short Time Fourier Transform (STFT) is

then used to transform the EEG signals to frequency domain.

Thus, only 5000 samples are available after the pre-processing

Basicemotionstimuli(IAPS)

Drivingsimulator

Feature Extraction(MFCC)

EEG Signals(Pre-

Processing)

EEG Signals(Raw Data)

Data Collection

Noise filter(Decimate)

Eyes Openand EyesClosed

RESULT

Precursor Emotion (MLP)

Testing(Eyes Closed Data)

Classification (MLP)

Training Emotion Data (VA)

Dat

a C

olle

ctio

n

Pre

-Pro

cess

ing

Feature Extraction

Classification

Fig. 4. Proposed research methodology block diagram for precursor emo-tion analysis.

3027


Sad

happy

Sad Sad Sad Sad

SadSad

happy

Fig. 5. Precursor emotion driver 1.

per channel per-emotion (one and half minutes sampling includ-

ing overlap).

100 features are extracted using the Kernel Density Estimation

(KDE) method for each channel (c3, c4, t3, t4) while 10 fea-

tures are extracted using the Mel-frequency Cepstral Coefficient

(MFCC) method. The KDE and MFCC are the two feature

extraction methods used in this study and analyzed. For KDE

total size of each channel will have 100 features per instances

per emotion yielding a total of 128 instances. Whereas 10 fea-

tures per instances per channel and 110 instances for MFCC

are obtained. The features extracted will be used as input to the

MLP classifier to train the network. The expected output during

training will be the valence and arousal values determined by

the affective space model (ASM). Finally, the results in valence

arousal approach are produced to get the degree of accuracy for

each driver. In this section, the MFCC data size is chosen as the

example in each experimental research method.

5. RESEARCH METHODOLOGYIn this research methodology, it comprises of noise filter method,

feature extraction method using Mel-frequency Cepstral Coef-

ficient (MFCC) and Kernel Density Estimation (KDE). Finally,

we have classification method by using Multilayer perceptron

(MLP).

5.1. Noise Filter Method (Decimate)

In order to filter the input data with low pass filter, decima-

tion process is then performed. The Matlab code; y = decimate(input, 3), has been used to reduce the sample rate of input by a

factor of 3. The decimated vector y is 3 times shorter in length

than the input vector. Thus, the size of emotion data which is

150000 has been decreased to 50000 samples after decimation

phase.

Happy CalmFear Sad

Sad

Happy Happy

Happy


FearSad


5.2. Feature Extraction (MFCC)

In this research study, feature extraction needs to be performed

on a non-speech signal. As such, we will be exploring on using

MFCC to extract this low frequency signal. 10 MFCC coeffi-

cients are used to capture the features.13 40 features are obtained

for the classification with 256 nfft, 83 sampling frequency and

20% overlap.13 The total number of features for all channels is

440 with 40 instances has been produced for the MFCC’s feature

extraction. Figure 4 illustrates the data of emotion that has been

extracted by using MFCC.

5.3. Classification (MLP)

In order to classify the extracted feature, MLP was adopted to

identify the precursor emotion. No of hidden layer is 1, while

number of neuron in the hidden layer is 10. The activation func-

tion for hidden layer is tan-sig and activation for output layer is

purelin. Learning rate is 0.01 and mean square error goal is 0.1.

6. RESULTSIn this experiment, eyes closed data has been analyzed in order

to identify the precursor emotions. The significant of this find-

ing is to have a better understanding precursor emotion effects

towards the driver behaviour. Eye closed has been consumed as

the precursor emotional state that the driver has before they were

invoke by any task. Therefore, it is important to know what kind

of emotion that the drivers have before they start the task.

Precursor emotion for driver 1 is a combination sad and happy.

This means the driver is having sad emotion in the beginning and

end with happy emotion. Thus, this emotion may influence the

driver behaviour. This is shown in Figure 5.

In Figure 6, it depicts the combination of happy, calm, fear

and sad emotion. The driver has happy emotion in the beginning,

having fear, sad and calm emotion in the middle and finally the

driver is having a happy emotion.

In contrast with driver 2, driver 6 has sad and fear emotion in

precursor emotion. This is shown in Figure 7. Thus, this emotion

may influence driver behaviour while the driver is driving the

car.

7. CONCLUSIONSIn this precursor emotion finding, it has been exemplified that

the emotions appears in eyes closed analysis is same as the other

analysis. The highest emotion that has been achieved has been

presented in the eyes closed analysis. Within 37 minutes, the pre-

cursor emotion has influenced the emotion for the rest analysis.

3028


Hence, it affects the driver behaviour especially in controlling

the driving simulator. Consequently, the precursor emotions have

the major influence in human emotion since it has been existed

in memory since long time.

Acknowledgments: This work is supported by the

grant (Ref: FRGS/2/2013/ICT07/UIAM/01/1) from the Ministry

of Education, Malaysia.

References and Notes1. S. Soedhodho, International Association of Traffic and Safety Sciences, IATSS

Research 33, 122 (2009).2. D. E. Clark, The affective reasoner: A process model of emotions in a

multi-agent system. Unpublished Doctoral Thesis, Northwestern University,Chicago (1992).

3. T. Bellet, B. Bailly-Asuni, P. Mayenobe, and A. Banet, Safety Science, Elsevier47, 1205 (2009).

4. N. Kamaruddin and A. Wahab, Driver behaviour analysis through speechemotion understanding, Proceeding of the 2010 IEEE Intelligent Vehicle Sym-posium (IV 2010) (2010), pp. 238–243.

5. A. Mehrabian and J. A. Russell, An approach to environmental psychology,MIT Press, Cambridge, MA, USA (1874).

6. P. J. Lang, The three system approach to emotion, edited by N. Birbaumer,A. Ohman, and The Organization of Emotion, Hogrefe-Huber, Toronto (1993),pp. 18–30.

7. P. Jahankhani, V. Kodogiannis, and K. Revett, IEEE Transaction on ModernComputing JVA’06, 120 (2006).

8. A. Savran, K. Ciftci, G. Chanel, J. C. Mota, H. V. Luong, B. Sankur, L. Akarun,A. Caplier, and M. Rombaut, Emotion Detection in the Loop from Brain Sig-nals and Facial Images, Workshop on Multimodal Interfaces, eNTERFACE’06(2006), pp. 11–101.

9. L. A. Sroufe, Emotional Development: The Organization of Emotional Life inthe Early Years, Cambridge Studies in Social and Emotional Development,Cambridge University Press (1997), pp. 58–64.

10. A. S. AlMejrad, European Journal of Scientific Research 44, 640 (2010).11. G. Chanel, J. Kronegg, D. Granjean, and T. Pun, Emotion assess-

ment: Arousal evaluation using EEG’s and peripheral physiological signals,Proceedings International Workshop on Multimedia Content Representation,Classification and Security, Istanbul (2006), pp. 530–537.

12. E. R. Kandel, J. H. Schwartz, and T. M. Jessell, Principles of Neural Science,McGraw Hill (2000).

13. N. M. Nor, A. Wahab, N. Kamaruddin, and H. Majid, Post accident analysis ofdriver affection, 15th IEEE Symposium on Consumer Electronics, ISCE2011,Singapore (2011).

Received: 17 December 2014. Accepted: 8 February 2015.

3029




Vol. 21, 3030–3033, 2015

Optimized Walficsh-Bertoni Model for Path Loss

Prediction DTTV Propagation in Urban

Area of Southern Thailand

Pitak Keawbunsong∗, Pitchaya Supannakoon, and Sathaporn Promwong

Faculty of Engineering (Telecommunication Engineering), King’s Mongkut’s Institute of Technology Ladkrabang,Chalongkrung Rd., Bangkok, 10520, Thailand

This article presents the optimization of a Walficsh-Bertoni path loss model to be used in designing the DTTVpropagation in an urban area of southern Thailand through the data collection on the signal power that thenetwork operators of 4 channels broadcasting within the distance of 2.5–6.5 Km. in urban Hat Yai, SongklaProvince, an area of high density with buildings. A least square method is used for the optimization while anefficient indicator is through statistical values of root mean square error (RMSE) and relative error (RE). Theresult is a new Walficsh-Bertoni model that has RMSE value lesser than the original model whereas the REvalue is also closer to zero than the original models. Subsequently, the Walficsh-Bertoni model is more precisein a prediction, making it optimized for use in planning the network.

Keywords: DTTV Propagation, Walficsh-Bertoni Model, Least Square Method.

1. INTRODUCTIONThailand launched the broadcasting of DTTV on April 01, 2014

with DVB-T2 standard and is currently under a process of

installing a main and a sub stations1 and simultaneously plans on

a design of a gap filler station in an urban area that faces with a

dead spot caused from the environment such as high rise build-

ings and their density that obstruct the TV signal.2 An effective

planning for the gap filler that includes the ability to cover the

proper areas, the use of a suitable transmitted power and the cost

saving requires an accurate predicted path loss model for the area.

Obviously, the development of the predicted model for the areas

using a least square method has been studied and presented in

many researches.3�4

Walficsh-Bertoni path loss model was initiated on the basis

of the loss from the diffraction of waves in the high rise areas

as well as the study on the signal receiving between buildings

at the diffraction of the waves in various directions.5 Thus, this

becomes interesting to apply for use in a design of a network

gap filler station within an urban area.

This article presents an optimized Walficsh-Bertoni path loss

model in urban areas of southern Thailand through using a least

square method. The second part illustrates a Walficsh-Bertoni

model in details whereas the third explains the data collection

for the optimization while the forth demonstrates an optimization


process of a path loss model and the fifth explains the result,

finally, the sixth is the conclusion.

2. WALFICSH-BERTONI MODELWalficsh and Bertoni present a semi-empirical path loss model6

for use in the environment of high rises and density as the model

assumes that the height of the buildings is a function of uniform

distribution. If the distance between buildings is equal, the wave

propagation will turn in various directions, through the building

rows onto the received antenna, as Figure 1.

A Walficsh-Bertoni model consists of 3 parts of path loss: free

space loss (PLfs), diffraction loss from rooftops (PLrooftops) and

diffraction and scatter loss from rooftops down and from a street

(PLdown�7

PLfs =−10 log��/4�r�2 (1)

PLdown = ��1

2�2�HB −hm�(2)

PLrooftops =[

0�1

(sin

√d/�

0�03

)0�9]2

(3)

sin can be written in terms of a height of a transmitted

antenna �ht�, a building height (HB), a distance (R) and the equa-

tion is:

sin= �ht −HB�/R (4)



12

3

4ht

HB

dR

Fig. 1. Propagation geometry for the Walficsh-Bertoni model.

From (3), a new equation can be:

PLrooftops = 0�01

(ht −HB

0�03R

)1�8

�d/��0�9 (5)

The total loss is:

PLtotal = 5�51

32�4

�ht −HB�1�8�1d

0�9

�HB −hm�2

�21

E3�8(6)

The Eq. (6) can be spread with a unit of dB as:

PLtotal = 89�5−10 log�1d

0�9

�HB −hm�2+21 log fc

−18 log�ht −HB�+38 logR (7)

�1 can be obtained from:

�1 =√�d/2�2 + �HB −hm�

2 (8)

fc is a frequency with a unit of MHz, ht is the height of a trans-

mitted antenna with a unit of metre, HB is the height of the

building with a unit of metre, hm is the height of a received

antenna with a unit of metre, d is a distance between buildings

with a unit of metre and R is a distance between the transmitted

station and the received antenna with a unit of metre.

3. DATA COLLECTIONThe data of the DTTV broadcasting stations of Thailand are pro-

vided by TPBS, one of the network and infrastructure operators

and the data collection for analysis is conducted through a signal

measurement in urban Hat Yai, Songkla Province at a distance

of 2.5–6.5 Km. away from the station by using a Drive Test

function of DVB-T2 PROMAX; HD RANGER+ model which

consisted of GPS and USB drive for recording the details. The

function is set to record at every 2 second. The received antenna

is installed on top of the pickup with the height of 2 metre from

the ground. The commercial antenna is Omni-directional from

Fig. 2. Hat Yai streets, Songkla Province where the data are collected usinga drive test function (Google map, 2015).

Table I. Details of DTTV broadcasting at Songkla stations.

FrequencyChannel (MHz) Brand transmitter Model Power (KW)

CH26 514 HARRIS UAX-2000T2HE 1.3CH42 642 HARRIS UAX-2000T2HE 1.3CH46 674 NEC DTL-10/1R0S 1.0CH54 738 NEC DTL-30/1R4SD 1.3

SPECTRUM.CO.LTD, Korea of Omni-Saturn model with a gain

rate (Gr) of −3 dBi. The area that is selected for collecting the

signal value is on the street full of high rise buildings on both

sides and 20 Km/hr is the speed limit. The analysis is through

the use of a medium value of the street which is 24 m and an

average height of the building which is 10 m. Using Google Map,

Figure 2 illustrates the signal measurement route in the urban

area.

The provincial DTTV broadcasting station is located

on Mt. CorHong at the latitude of 7 0’ 57.95” and the longitude

of 100 31’ 12.17” with 366 m above sea level whereas the height

of the transmitted antenna is 66 m, as details in Table I.

The relation of received signal power (Pr) and the signal path

loss value (L) can be described as:

EIRP = Pt +Gt −Cl1 −Cl2 (9)

L�dB�= �EIRP+Gr�−Pr (10)

EIRP is an effective isotropic radiation power value using a unit

of dBm; Gt and Gr is a gain rate of the transmitted antenna

and the received antenna using a unit of dBi; Pt and Pr is the

DTTV transmitted power and the received signal power using a

unit of dBm; Cl1 and Cl2 are the transmitted loss value in the

combiner and the transmitted line with a unit of dB. The system

works by the broadcasting power of every channel sends through

a rigid line having a size of 1 5/8” into the combiner, Spinner;

CCS-6WAY, with a loss value (Cl1) of 0.55 dB and the output

from the combiner further sends through an RFS transmission

line, Flexwell HF 5”, with a total loss value (Cl2) of 1.186 dB

feeding into an antenna, Horizontal RFS; PHP48U, with the total

gain (Gt ) of 18.35 dBi.

4. OPTIMIZATIONThis study uses a least square method for an optimization process

to look for a parameter value in order to adjust a Walficsh-Bertoni

model to get closer to the measured data as much as possible.

The Eq. (7) is determined that:

E0 = 89�5+21 log fc (11)

Table II. The result of the constant parameter obtained from the

optimization.

Linear least Optimizedsquare parameters

Frequency a b E0new �sysnew

CH26 (514 MHz) 125.177 2.173 29.727 0.057CH42 (642 MHz) 117.622 3.222 22.172 0.085CH46 (674 MHz) 114.983 3.598 19.533 0.095CH54 (738 MHz) 113.399 4.036 17.949 0.106

3031


Table III. The result of the statistical comparison.

URMSE (dB) Relative errors

Frequency After Before After Before

CH26 (514 MHz) 14�17 18.22 0.094 0.126CH42 (642 MHz) 11�74 13.93 0.074 0.093CH46 (674 MHz) 10�79 12.86 0.065 0.084CH54 (738 MHz) 9�45 12.32 0.055 0.081

Esys =−10 log�1d

0�9

�HB −hm�2+18 log�ht −HB� (12)

�sys = 38 logR (13)

The least square method8 is a function of the sum when deviation

square becomes minimum.

P�a�b� c� � � ��=n∑

i=1

�yi−ER�xi�a�b� c� � � ��2 = min (14)

yi = value obtained from the measurement at the distance of xi.ER�xi�a�b� c� � � ��= the result from the predicted model at the

distance of xi.a, b, c = model parameter for the optimization.

n= the number of data from the measurement.

All partial differential results of P function should be equal to

zeros

�P/�a= 0� �P/�b = 0� �P/�c = 0 (15)

From (11)–(13) we can spread them to:

a= E0 +Esys� b = �sys (16)

When logR= x from (16), a new equation can be written as:

ER = a+bx (17)

From (17), we can look for a constant value of a, b from the

measured data group and the (17) can be spread into:

n∑i=1

(�yi−ER�xi�a�b�c��

�ER

�a

)=∑

�yi−a−bxi�·1=0 (18)

n∑i=1

(�yi−ER�xi�a�b�c��

�ER

�b

)=∑

�yi−a−bxi�·xi=0 (19)

2 2.5 3 3.5 4 4.5 5 5.5 6 6.5 780

90

100

110

120

130

140

150

160

Distance (Km)

Pat

h lo

ss (

dB)

CH26 (fc = 514 MHz)

Measurement path lossWalficsh-Bertoni modelOptimized Walficsh-Bertoni model

Fig. 3. The comparison of pre and post optimization of Walficsh-Bertonimodel at CH26 with the frequency of 514 MHz.

2 2.5 3 3.5 4 4.5 5 5.5 6 6.5 780

90

100

110

120

130

140

150

160

Distance (Km)

Pat

h lo

ss (

dB)

CH42 (fc = 642 MHz)



(18) and (19) can be re-spread into:

n ·a+b∑

xi =∑

yi� a∑

xi +b∑

x2i =

∑�xiyi� (20)

Replace the variable a�b in (20) and we get a statistical estimated

value of the parameter a�b as:

a=∑x2i

∑yi−∑

xi∑xiyi

n∑x2i −�

∑xi�

2� b= n

∑xiyi−∑

xi∑yi

n∑x2i −�

∑xi�

2(21)

From (21), we can get the value a�b of the path loss data group

from the measurement and from (7) and (16), we get the value

of the offset and slop of the original Walficsh-Bertoni path loss

model, as:

E0new = a−Esys� �sysnew = b

38(22)

5. FINDINGS ANALYSISThe optimization of the Walficsh-Bertoni model to measure the

efficiency is done by looking for a statistic value of RMSE and

2 2.5 3 3.5 4 4.5 5 5.5 6 6.5 780

90

100

110

120

130

140

150

160

Distance (Km)

Pat

h lo

ss (

dB)

CH46 (fc = 674 MHz)



3032


2 2.5 3 3.5 4 4.5 5 5.5 6 6.5 780

90

100

110

120

130

140

150

160

Distance (Km)

Pat

h lo

ss (

dB)

CH54 (fc = 738 MHz)



RE from the measured signal power value which can be found

from the following equations.9

RMSE =√∑

�PLm −PLapprox�2

N −1(23)

RE = �PLm −PLapprox��PLm�

(24)

PLm is the measured path loss value; PLapprox is the Walficsh-

Bertoni path loss value; N is the amount of measured data where

the amount of 2300 points of each channel is selected for this

research.

The optimized result of the Walficsh-Bertoni model is the con-

stant parameter a�b, E0new and �sysnew as shown in Table II.

It is noticeable that the value of E0new decreases when the fre-

quency increases whereas �sysnew increases when the frequency

increases. The statistical comparison shown in Table III indicates

that the new Walficsh-Bertoni model is more accurate than the

original one. This is because, after the optimization, the values

of RMSE and RE reduce and they reduce when the frequency

increases. Figures 3–6 compare the pre and post optimization of

the frequency of each channel. It can be observed from the curve

line that in Figure 3; CH26; 514 MHz, the post optimization of

Walficsh-Bertoni path loss model is slightly close to the mea-

sured data whereas in Figure 4; CH42; 642 MHz, the curve line is

closer to the measured data than the one of Figure 3. In Figure 5;

CH46; 674 MHz, the curve line is closer to the measured data

than the one of Figure 4 and in Figure 6; CH54; 738 MHz, the

curve line is again closer to the measured data than the one of

Figure 5. This is consistent with the statistic result showing that

the new model are closer to the measured data that are higher

than the original one and that the new Walficsh-Bertoni model of

the high frequency is closer to the measured data than the ones

of low frequency.

6. CONCLUSIONThe result of Walficsh-Bertoni model obtained from the opti-

mization shows the suitability for use in designing a gap filler

station of the DTTV propagation in the urban Hat Yai, Songkla

Province where the dead spot exists. The model can also be

used as a functional reference for the urban areas in the south

of Thailand in order to appropriately cover the areas. The abil-

ity to use the right broadcasting power leads to cost reduc-

tion. The statistical efficiency indicator of each signal frequency

shown in Table IV reveals the decreasing value of RMSE and RE

upon being compared with the original models prior to the opti-

mization as the value decreases when the frequency increases.

Figures 3–6 demonstrate a clear result in which the Walficsh-

Bertoni model is closer to the measured data than the origi-

nal models. The research findings on Walficsh-Bertoni path loss

model optimization demonstrate the frequency variations that are

closer to the measured data. Seeking for the optimized upper

and lower boundaries of the path loss model through a confident

interval estimation method in order to create the more precise

prediction can be a significant future study topic.

References and Notes1. Office of The National Broadcasting and Telecommunication Commission, The

Radio Frequency for Digital Terrestrial Television in Thailand, NBTC, December(2012).

2. S. R. M. de Carvalho, Y. Iano, and Rangel, ISDB-TB field trials and cover-age measurements with gap-filler in suburban environments, IEEE InternationalSymposium on Broadband Multimedia Systems and Broadcasting, Nuremberg,Germany, June (2011).

3. J. Chebil, A. K. Lwas, Md. R. Islam, and Al-Hareth Zyoud, Proc. CSIT 5, 252(2011).

4. A. Bhuvaneshwari, R. Hemalatha, and T. Satyasavithri, Statistical tuning ofthe best suited prediction model for measurements made in Hyderabad cityof Southern India, Proceedings of the World Congress on Engineering andComputer Science, San Francisco, USA, October (2013).

5. I. Joseph and B. Michael, American Journal of Physics and Applications 30, 10(2013).

6. J. Walfisch and H. L. Bertoni, IEEE Transection on Antenna and Propagation36, 1788 (1998).

7. J. Isabona and S. Azi, International Journal of Engineering and InnovativeTechnology 2, 14 (2012).

8. B. S. L. Castro, M. R. Pinheiro, and G. P. S. Cavalcante, Optoeletronics andElectromagnetic Applications 10, 106 (2011).

9. R. Mardeni and K. F. Kwan, Progress in Electromagnetics Research 13, 91(2010).

Received: 30 September 2014. Accepted: 3 November 2014.

3033




Vol. 21, 3051–3054, 2015

Hermite-Gaussian Mode Division Multiplexing for

Free-Space Optical Interconnects

Angela Amphawan1�2�∗ , Sushank Chaudhary1, and Tse-Kian Neo3

1Integrated Optics Group, School of Computing, Universiti Utara Malaysia, Sintok, Kedah, Malaysia2Research Laboratory of Electronics, Massachusetts Institute of Technology, Cambridge, Massachusetts, 02139, USA

3Faculty of Creative Multimedia, Multimedia University, Cyberjaya, Selangor, 63100, Malaysia

In keeping pace with the exponential growth in network traffic, it is imperative to explore new multiplexing degreesof freedom in addition to wavelength, time, intensity and phase. The next frontier in free-space optics (FSO)envisages the eigenmode of an optical resonator as a new multiplexing degree of freedom. This paper focuseson mode division multiplexing of two parallel 2.5 Gbps channels on spiral-phased Hermite-Gaussian modesHG 10 and HG 11 for free-space optical interconnects at a wavelength of 850 nm. The signal-to-noise ratios,eye diagrams and modal analyses are investigated.

Keywords: Mode Division Multiplexing (MDM), Free-Space Optics (FSO), Optical Interconnects, Hermite-Gaussian Mode, Spiral Phase, Vortex Lens, Data Center.

1. INTRODUCTIONFree-space optics (FSO) technology transpires from prevailing

wireless communications and optical fiber communications. FSO

uses a high-speed optical carrier to enhance transmission capacity

without an optical fiber.1 FSO is appropriate for indoor appli-

cations such as losses due to scattering, absorption, turbulence

and scintillations indoors are constrained in relation to the atmo-

spheric noise level outdoors.2 A pioneer application for indoor

FSO is remote patient monitoring for vital signs such as breath-

ing, blood coagulation, and heart rate.3�4

Most FSO systems adopt wavelength division multiplexing

(WDM) to enhance channel capacity.5�6 Although WDM has

been the workhorse for several decades, bandwidth-rich applica-

tions, proliferation of smart mobile devices and device-to-device

connections have increased network traffic significantly.7�8 In

keeping pace with the escalating traffic growth, it is imperious to

explore new multiplexing degrees of freedom for increasing the

channel capacity.

The next frontier in FSO envisages the optical eigenmode of

an optical resonator as a new multiplexing degree of freedom. In

mode division multiplexing (MDM) optical eigenmodes are used

to drive the propagation of a been made by various mode gener-

ation mechanisms number of independent channels. Appreciable

progress has for MDM for optical fiber communications such

as spatial light modulators,9 optical signal processing,10–13 few


mode fiber14�15 and photonic crystal fibers16 to provide distinct

data channels and prevent modal crosstalk.

Nevertheless, MDM is still at its infancy in FSO and a

few types of mode profiles are being explored for prospective

channels in FSO MDM. In Refs. [17, 18], the performance of

MDM of orbital angular momentum (OAM) modes from dif-

ferent spatial light modulator encodings was evaluated through

free-space turbulence under various atmospheric turbulences and

system penalty. Hermite-Gaussian modes HG 00 and HG 01

were used for MDM in a 2.5 Gbps–20 GHz FSO link.28 In

Ref. [19], expressions for average intensity and effective size

of Laguerre-Gaussian (LG) and Bessel-Gaussian Schell-model

(BGSM) beams are derived to analyze the effects of turbulence

and laser coherence on the beam profile. In Ref. [20], analyt-

ical expressions related to the Wigner distribution function of

a Laguerre-Gaussian Schell model (LGSM) beam in turbulent

atmosphere are computed to investigate statistical properties of

the coherence and propagation factor. In Ref. [21], the spec-

tral density of cosine Gaussian-correlated Schell-model (CGSM)

beams diffracted by an aperture is derived to investigate the effect

of the spectral density distribution of CGSM beams.

In this paper, instead of using pure HG modes, we intro-

duce two spiral-phased HG modes for FSO MDM using vortex

lenses and analyze the modal decomposition. The remainder of

the paper is organized as follows. Section 2 elucidates the main

features of the MDM model and simulation parameters. Section 3

describes the results and discussions, followed by the conclusion

in Section 4.



Fig. 1. MDM of spiral-phased HG 10 and HG 11 modes for indoor FSO link.

2. SYSTEM DESCRIPTIONFigure 1 shows a block diagram of proposed FSO model using

spiral-phased HG modes designed in OptiSystem software.22 As

shown in Figure 1, the proposed architecture consists of two

independent non-return-to-zero (NRZ) encoded channels, each

carrying 2.5 Gbps data stream over an 850 nm optical spatial

carrier. Two laser modes of spot size 5 �m are used for data

transmission: HG 1, 0 for Channel 1 and HG 1, 1 for Channel 2.

The HG mode is described mathematically as:23

�m�n�r�� = Hm

(√2x

wo�x

)exp

(− x2

w2o x

)exp

(j�x2

�Rox

)

Hn

(√2y

wo�y

)exp

(− y2

w2o y

)(j�y2

�Roy

)(1)

A vortex lens is used to apply a spiral phase transformation

to each HG mode generated as shown in Figure 2. The applied

phase is given by the following:

T �x�y� = exp

[−�n�x2 +y2�

2�f+m tan−1

(x

y

)](2)

where f is the focal length of the lens, m is the vortex index

and n is the refractive index. For Channel 1, the vortex lens has

vortex index, m= 2 whereas for Channel 2, m= 5. The transverse

electric field used for the two modes used in Channel 1 and

Channel 2 are depicted in Figure 2. The output of two channels

are combined, amplified and transmitted through free-space of

varying distances from 200 m to 1000 m. The link is free from

atmospheric turbulences and suited for indoor applications. The

link equation for free space optics29 is modelled by:

PReceived = PTransmitted

(d2RR

�dT +�R�2

)10−�R/2 (3)

where dR defines receiver aperture diameter, dT is the transmitter

aperture diameter, � is the beam divergence, R is the range and

� is the atmospheric attenuation. At the receiver side, the trans-

mitted mode is extracted based on non-interferometric modal

decomposition.30 The output mode is then fed to a spatial PIN

detector followed by low-pass Gaussian filter to retrieve the orig-

inal baseband signal.

3. RESULTS AND DISCUSSIONIn this section, results from our proposed MDM based FSO link

are presented and discussed. The signal-to-noise ratio (SNR) and

total received power at the receiver are shown in Figures 3 and 4

respectively. Both SNR and total received power graphs demon-

strate that Channel 1 carrying spiral-phased HG 10 performs

better than Channel 2 carrying spiral-phased HG 11. The SNR

deteriorates with FSO range for Channel 1 with SNR values of

34.11 dB, 19.36 dB and 9.19 dB for Channel 1 for a FSO link

of 200 m, 600 m and 1000 m respectively. The SNR for Chan-

nel 2 declines with a slightly steeper slope, with SNR values of

30.52 dB, 10.11 dB and 2.32 dB for a FSO range of 200 m,

(a)

(b)

Fig. 2. Excited modes (a) HG 10 mode with vortex index, m = 1 forchannel 1 (b) HG 11 mode with vortex index, m = 3 for channel 2.

3052


Fig. 3. SNR versus FSO range.

600 m and 1000 m respectively. On the other hand, the total

power received for Channel 1 are −65.43 dB m, −80.31 dB m

and −90.14 dB m whereas for Channel 2, the total received

power is −69.22 dB m, −89.18 dB m and −97.32 dB m for a

FSO link of 200 m, 600 m and 1000 m respectively.

The modal content at the receiver is computed in terms of

power coupling coefficient of linearly polarized (LP) modes

using noninterferometric modal decomposition,19 arranged in the

order of descending power coupling coefficient, as shown in

Figure 5. For Channel 1, the power is coupled predominantly into

modes with azimuthal mode order of 2, with the largest power

coupled into mode LP 0,1, followed by LP 2,1; LP 2,2; LP 0,2

and LP 0,3. For Channel 2, the power is coupled predominantly

into modes with an azimuthal mode order of 5, with the largest

power coupled into mode LP 1,1 followed by LP 5,1; LP 5,2;

LP 5,3; LP 5,4 and LP 4,7. This agrees with the spatial profiles

of the spiral-phased HG modes. Wide eye openings are attained

for both channels are shown in Figure 5, which confirm success-

ful data transmission of 2× 2�5 Gbps over a FSO link of 600

meters for Channel 1 and 400 meters with Channel with accept-

able SNR.

Fig. 4. Total received power against FSO range.

(a)

(b)

Fig. 5. Modal decomposition at the receiver in terms of descending orderof power coupling coefficients in linearly polarized modes: (a) Chanel 1(b) Channel 2.

4. CONCLUSIONIn this work, 2 × 2�5 Gbps data transmission is realized for a

400 m indoor FSO link by MDM of two independent channels on

spiral-phased HG 10 and HG 11 modes. The results reveal that

Channel 1 propagating spiral-phased HG 1,1 mode with vortex

index, m = 1 is more robust than Channel 2 propagating spiral-

phased HG 0,1 mode with vortex, m = 3. The model may find

applications in optical interconnects in mega data centers.

References and Notes1. S. Chaudhary and A. Amphawan, Journal of Optical Communications 35, 327

(2013).2. A. K. Majumdar and J. C. Ricklin, Free-Space Laser Communications: Princi-

ples and Advances, Springer, New York (2008), Vol. 2.3. A. M. Khalid, G. Cossu, and E. Ciaramella, Diffuse IR-optical wireless system

demonstration for mobile patient monitoring in hospitals, 2013 15th Interna-tional Conference on Transparent Optical Networks (ICTON) (2013), pp. 1–4.

4. L. Chevalier, S. Sahuguede, and A. Julien-Vergonjanne, Investigation of obsta-cle effect on wireless optical on-body communication performance, 2014 21stInternational Conference on Telecommunications (ICT) (2014), pp. 103–107.

5. Y. Wang, X. Huang, L. Tao, and N. Chi, 1.8-Gb/s WDM visible light com-munication over 50-meter outdoor free space transmission employing CAPmodulation and receiver diversity technology, Optical Fiber CommunicationConference, Los Angeles, California, (2015), p. M2F.2.

6. A. Aladeloba, M. Woolfson, and A. Phillips, J. Opt. Commun. Netw. 5 (2013).

3053


7. (2015, Feb. 3, 2015) Cisco Visual Networking Index: Global Mobile Data Traf-fic Forecast Update 2014–2019 White Paper.

8. K. Nisar, A. Amphawan, and S. B. Hassan, Int. Journal of Advanced Pervasiveand Ubiquitous Computing (IJAPUC) 3, 50 (2011).

9. !!! INVALID CITATION !!!10. S. O. Arik, D. Askarov, and J. M. Kahn, J. Lightwave Technol. 32, 1841

(2014).11. S. O. Arik, J. M. Kahn, and K.-P. Ho, IEEE Signal Processing Mag. 31, 25

(2014).12. A. Amphawan, Optics Exp. 19, 23085 (2011).13. A. Amphawan, V. Mishra, and K. N. B. Nedniyom, J. Mod. Opt. 50, 1745

(2012).14. C. P. Tsekrekos and D. Syvridis, IEEE Photon. Technol. Lett. 24, 1638 (2012).15. Y. Jung, R. Chen, R. Ismaeel, G. Brambilla, S. U. Alam, I. P. Giles, et al.,

Optics Express 21, 24326 (2013).

16. A. Amphawan and N. M. A. A. Samman, Tiering effect of solid-core pho-tonic crystal fiber on controlled coupling into multimode fiber, SPIE OpticalEngineering+Applications: Photonic Fiber and Crystal Devices: Advances inMaterials and Innovations in Device Applications VII, San Diego (2013).

17. J. Wang, J.-Y. Yang, I. M. Fazal, N. Ahmed, Y. Yan, H. Huang, et al., Nat.Photon. 6, 488 (2012).

18. Y. Ren, H. Huang, G. Xie, N. Ahmed, Y. Yan, B. I. Erkmen, et al., OpticsLetters 38, 4062 (2013).

19. J. Cang, P. Xiu, and X. Liu, Optics and Laser Technology 54, 35 (2013).20. R. Chen, L. Liu, S. Zhu, G. Wu, F. Wang, and Y. Cai, Optics Express 22, 1871

(2014).21. L. Pan, C. Ding, and H. Wang, Optics Express 22, 11670 (2014).22. Optiwave, Optisystem, edited by Ottawa, Canada (2014).23. J. Enderlein and F. Pampaloni, Journal of the Optical Society of America A

21, 1553 (2004).

Received: 23 January 2015. Accepted: 18 April 2015.

3054




Vol. 21, 3060–3064, 2015

Analysis of Data Quality Maturity in a

Higher Education Institution

W. Yohana Dewi Lulu1�2�∗, Adhistya Erna Permanasari1, Ridi Ferdiana1, and Lukito Edi Nugroho1

1Electrical Engineering and Information Technology, Gadjah Mada University, Yogyakarta 55281, Indonesia2Information System, Polytechnic Caltex Riau, Pekanbaru 28265, Indonesia

The organizations use data as an important asset. Data became the important assets in some organizationsmay give effect due to the cost. Data quality standards for non-profit-oriented education organizations have notspecifically defined. This study aims to provide guidance for non-profit organizations in order to produce qualitydata. The focus of analysis is a data modeling, determination of dimensions of quality and maturity level ofdata quality in organizations. The first step needs to be done is to create a data quality model based on entityrelationship data and to fix dimensions of data quality to be used. Dimensions are used to follow the intrinsic,contextual, representative, accessibility data quality. Determination of the level of maturity of the initial and finalof data quality analysis process used to make recommendations next stage. The results showed that stages ofdefining quality varied by the user complicate the process of preparing the data model quality. Maturity of dataquality management were mostly at the defined data management (level 2) and the process move to level 3.Maturity data quality management requires analysis related to data and process. Evolving business processorganization should be aligned with the data quality standards have been set.

Keywords: Data Quality, Maturity of Data Quality, Higher Education.

1. INTRODUCTIONData are very useful in an organization. Data can be an asset to

the organization. An understanding of the functions and bene-

fits of the data in general, should be shared by all parts of the

organization. The data would be useful if data are presented in

accordance to the needs of their owners. Usefulness of data varies

according to the needs of their owners. This causes differences

in the benefits of data for every data owner.

The use of data in many organizations is used to basis for

decision-making to carry out operational and strategic level.1–3

Data quality is a critical factor for achieving strategic and oper-

ational business.4 Poor data quality will have a negative impact

in many ways in organizations.5 If the data is of poor quality

but it did not be identified and corrected, it could have economic

and social impacts significantly negatively on an organization.

Efforts to improve the quality of data for a particular purpose or

long-term goals require improvements in the quality of data that

involves a data-driven and process-driven.5–7

Knowledge about data’s purpose right from the beginning of

process can be contribute to increase data quality.8 There is one

perspective that can be used to analyze and improve data quality.5

This paper provide a comprehensive process for exploration data


quality maturity. The purpose of this paper is to describe an

understanding of the quality of data within the organization with

the concept of maturity. This research was conducted by study-

ing previous research literature and analysis of case studies at a

small higher education.

The result show benefit and limitations of the approach, allow-

ing practitioner to tailor the approach to their needs. Analysis

of process driven that used in business processes for marketing

process in the case study, demonstrating the benefits, a sense of

ownership and a better understanding of the quality of data.

The remainder of the paper is structured as follow Section 2

introduces the basic theory of data quality and maturity and also

basic theory of business process. Section 3 analyse case study,

Section 4 report the result and finally Section 5 presents the

Conclusion.

2. BASIC THEORY OF DATA QUALITY,

MATURITY AND BUSINESS PROCESSThe main rationale underlying this paper is to achieve an under-

standing of the quality of data within the organization with the

concept of maturity. In this section we will explain the concept

of data quality, business process management and business pro-

cess management maturity, based on the study of literature that

has been done.



2.1. Data Quality

The quality of the existing data is very important for an orga-

nization. The definition of quality varies greatly various organi-

zations and academia. The International Standard Organization

(ISO) defines quality as degree to which a set of inherent char-

acteristics fulfils requirements. A collection of data is of high

quality, in customers view, if the data meet his, her or its needs.9

According to Genichi Taguchi,10 quality is closeness to target and

deviations means loss to society. However, J. Juran and American

Society of Quality Control referred quality as fitness for use11

and fitness is defined by customers.12–14 Organization study qual-

ity of data based on the definition of fitness for use and fitness

is customers satisfaction for their purposes.

Data quality factors can be divided into three domain: data

service, data value, data structure quality. Figure 1 illustrates a

recurring stage of evaluating the quality of the data from the

three elements (structure, value and service). Management of data

quality includes the level of depth and greater use of quality

components.

The greater use of data quality covering the integration of

information systems within the scope of the organization’s

database. Greater use of the data quality including quality lev-

els defined data structures and process integration as well as the

overall consistency of the data.15 However, it is hard to evalu-

ate data quality objectively without considering the width of data

under the discrete and complicated information system enviro-

ment data quality can be assessed from the standpoint of the

product. Therefore, the data has producer, consumers and brokers

in the process of fulfilling the data itself.14

Data is a collaboration of two components, namely the data

model and the data value.16 The data model describes the con-

dition of the real world with structured. Entity data model

expressed as objects, attributes and ideas in the real world.

Attributes in the data model involves the nature of entities and

relationships between entities. Value of data provides information

realization of an attribute in the entity or association. The concept

of data quality is the same thing with the quality of the product.

The quality of the product produced depends on the design and

production process itself. Understanding the meaning of quality

and how quality is measured will assist in the design process of

good quality. The quality of data that involves many factors and

components referred to as a multidimensional concept.17�18

2.2. Data Quality Assesment

Data quality assessment can provide an idea of the level of qual-

ity of the existing data and establish appropriate criteria for data

quality improvement process further. In their improvement steps,

methodologies adopt two general types of strategies, namely

data-driven and process-driven.6�19 Data-driven strategy can help

the process of improving the quality of data with making changes

and updates to the value of the data. Process-driven strategies can

help to improve the quality of data to make changes and modify

the data creation process.2�6 As an example, a process can be

Fig. 1. Data quality model.15

redesigned by including an activity that controls the format of

data before storage.

Process-driven data quality management aims to identify root

cause of poor data quaility. Process-driven conducts process con-

trol or redesign activities. In process driven, process modeling is

providing the means to understood and commicate.6 Data gener-

ated in the right business processes that are expected to have the

right quality. Many organizations often forget this, organizations

pay more attention to the data resulting from the process of how

to get it.

Data quality methodology apply the concept of data-driven

and process-driven as one of the main strategies in activities

to improve the quality. Simply put, a data-driven strategy can

improve data quality by directly modifying the data values, and

strategy driven process of improving quality by redesigning pro-

cesses that will create or modify the process of inputting the

data and data processing. Some methodologies use strategies,

phases, activities and different quality dimensions, but the stages

of assessment and improvement has always been a part that is

not abandoned.

For the assessment phase, the diagnostic process data quality

and relevant quality dimensions, using adequate data quality tools

(DQT). The increase mainly focus:

(a) The process of identifying the root of the error.

(b) Make improvements of error with the right DQT and

(c) Make changes to the design of specific DQ engineering and

redesign of data creation process to improve its quality.

Batini2 presents and compares some of the most widesread

methodologies.

2.3. Data Quality Management Maturity

Data quality evaluation and management should be defined

from many points of view such as total corporate integration

management,20–22 data structure quality management, and man-

agement maturity stages. The data management maturity model

that is used as the capability maturity model of software pro-

cess evaluation (Fig. 2).15�22�23 According to Ryu,15 he proposed

maturity stage that consists of four levels, namely initial, defined,

managed and optimized.

Level 1. Initial (Data Management Stage). This stage defines the

structure of the quality of the data and rules used in the database

by management.

Level 2. Defined data management. This stage deals with mod-

eling of data models, both logically and physically. This phase

establishes the model data according to pre-established database.

Level 3. Managed. These stages are managed using standard

enterprise. Management of metadata at this stage includes the

entire enterprise data. The integration process is developed at this

stage.

Level 4. Optimized. Stages set management in the form of archi-

tectural models. This stage optimizes data management, data

models and relationships within the company standards.

2.4. Basic Theory Business Process

The process is a stage arranged in a logical work in creat-

ing goods and services, including the creation of the use of

resources.24�25 Process-driven data quality management is used

to find the root cause of data quality is low.20�26 In contrast with

data correction, process defects repaired and adjusted to main-

tain improvement. Modeling process should provide a means to

3061


Fig. 2. Data quality management maturity model.15

understand and communicate the process to be able to control

and redesign activities.2 Quality management based data-driven

process is not limited in the domain of information systems.

Process-driven data quality needs to be supported by other factors

that may be related, the definition of data quality and suitability

of the dimensions used.

Concerning process-driven activities, a typical business pro-

cess reengineering activity,27 for a comprehensive discussion is

composed of three activities: Mapping and analyzing the as-is

process; Designing the to be process, according to one or multi-

ple alternatives; Implementing the reengineered process and con-

tinuously improve. Business processes become key elements of

success of an organization.

Business process describes how the organization operated and

how the role of the organization’s performance. Organizations

need to strive to get better performance. Therefore organizations

need to understand the maturity of their business processes.28�29

Business Process Management (BPM) is concerned with the effi-

cient management of business processes and their continuous

improvement.30 For managing processes, BPM provides a set of

structured methods and technologies.31 Current research provides

an overview of the several BPM standards across the BPM life

cycle.32 Several BPM maturity models exist to assess organiza-

tions maturity and provide guidance for its improvement.

3. ANALYSIS OF CASEPolytechnic Caltex Riau (PCR) was built by PT. Caltex Pacific

Indonesia (CPI), BP Migas and the Government of Riau

Province, as proof of their commitment to improving the quality

of human resources in Riau. PCR aims to produce quality human

resources in the field of applied technology, which has the knowl-

edge and integrity. The resulting human resources are expected

to meet the manpower needs in the industry both nationally and

internationally.

PCR is trying to focus on the process of his own efforts to

achieve the vision and mission. One of the very early process

of note is the new admissions process. This process is the input

to the production process. Since four years ago, new admis-

sions process had to use information technology as a marketing

medium. Then, marketing team began to realize the role of the

data obtained in the previous year is very useful for the next

admissions process.

PCR required to be more competitive advantage to the other.

Good data management will greatly assist the higher education in

many ways. Data quality in higher education need to be examined

more carefully to support various processes in higher education,

such as the process of marketing, learning data processing and

other learning outcomes.

At this time PCR as one of higher education that focuses on

the skills of its graduates have more than 10 years old. This

year the number of active students who are studying in the PCR

reached 1600 student. Since its inception, PCR-based managed

IT concepts. Processing of the data generated by the share in the

PCR process has been managed IT, although not achieving the

integration process.

Maturity level assessment of data quality management will be

done by providing a preliminary question. Assessment given to

the users of the application for 60 user PCR. The results of the

assessment have been distributed only returned 39 pieces. The

data can be validated and accepted in the processing of as many

as 37.

Based on the results in Figure 3 Level satisfied data quality

management shows that the management of the quality of exist-

ing data has reached 73% level of satisfaction. This result does

not give a picture directly to the successful management of data

quality.

Therefore conducted direct interviews with five users are lim-

ited by user applications with the same end goal. Based on the

results of interviews conducted, showed that the level of satis-

faction about the quality of the data that is otherwise still based

on the viewpoint of the work that has been defined for them.

The absence of standards managed organization as a whole in

the process of data quality.

Data quality dimensions assessed include four aspects of data

quality. Intrinsic aspect of assessing the dimension accuracy and

believability. Contextual aspects of the quality of the data that

already includes the process of software engineering require-

ments only assess the dimensions of completeness. It is based

that the data provide the best quality if according to its users.

In the aspect of data quality assessed dimensional representation

of consistency and duplicate, whereas the accessibility aspects of

the assessment on both dimensions.

Determination of dimensions which are valued at the ini-

tial survey and interview data users within the scope of the

Fig. 3. Level satisfied data quality management.

3062


application. Survey and initial interviews covering all aspects of

data quality version of Wang.18 Assessment is done by using

the interval 1–9 to provide flexibility assessment. Results of the

importance of data quality dimensions are given in Figure 4 Level

Important Data Quality Dimension.

In 4 Level Important Data Quality Dimension it can be seen

that the component dimensions have the highest level of interest

is the believability and then the dimensions of access security.

Level of understanding of leaders and managers of higher educa-

tion on the concept of quality management data is done with the

interview. It is the focus of the interview, relating to the design

process associated with the data, data storage and data retrieval.

Components of data quality are an indicator for each data collec-

tion process. Data collection was done through interviews with

the leaders of the organization.

Clarification of data quality aspects define earlier, clarified and

adapted to the existing file governance, management, marketing

staff and IT staff. Overall respondents were given a brief under-

standing of the concepts and data quality management maturity.

This is used to provide the same level of understanding of the

quality of the data.

To date, this institusion do not have a company data guide-

lines nor a management data center initiative. For example, they

still use the same attribute with different modifications. They had

difficulty in defining the marketing data and perform risk man-

agement and control it manually.

Data used and created in times of need in the process. Associ-

ated with the process used to identify the data used, the necessary

Awareness, Exploration, Reporting, Fixing and Preventing. Man-

agement believes that root cause analysis is required from the

beginning until today to be able to fix the attribute data needed

in the marketing process.33

4. RESULTThe main driver for PCR to undertake the data quality manage-

ment initiative was costumer data, which is in line with the most

relevant areas for increase the marketing process and to maintain

the customer. By applying the data quality framework. PCR can

have began bottom-up, being only focused on a particular dataset

and not having a organization data policy, strategy or governance.

Utility driven approach can be used to better focus on specific

database.34

Results of analysis and interviews conducted on PCR showed

that PCR is still at the stage of defining data management

Fig. 4. Level important data quality dimension.

(level 2). This is evident because they are still improving business

processes and choosing new tools for data management. The next

stage they are going is the stage to prevent data quality problems

and must make centralized management. Management standards,

rules and classifications for all business processes.

The concept of intrinsic DQ gives meaning that the data has

a quality of its own point of view. DQ contextual concept is

discussed that the data quality requirements must be taken into

consideration in the process of how the data is obtained and

used. Representation and accessibility DQ how do system focuses

more on the quality of data. All of this concept provides an

understanding that high quality data must be intrinsically good,

contextually appropriate to the task, clearly represented, and can

be accessed by the consumer data.

5. CONCLUSIONSData quality management maturity begins in the standard-setting

process the desired data quality user. Maturity management data

quality requires analysis of the relationship between the data and

processes. Developments in the organization’s business processes

must be aligned with the data quality standards that have been

set. This also applies vice versa, where the standard of data qual-

ity can be increased or changes with the organization’s business

processes.

In our case study in PCR, the level of maturity of the quality

of the data is still in the understanding of data quality standards.

If no explanation at the beginning of the process of data qual-

ity, better understanding of data quality processes that already

exist and of all the different users. Further development can be

done by involving more components of organizational business

processes to gain a broader picture and depth data quality man-

agement maturity. Further assessment of the various viewpoints,

such as the application point of view, and private organizations

or user standpoint, would be required. Maturity assessment needs

to consider elements of the user. How, who and what the role of

the user in the process of an organization’s data quality. How and

to what extent a user to understand the level of need for quality

data.

References and Notes1. M. Eppler and M. Helfert, A classification and analysis of data quality costs,

International Conference on Information Quality (2004).2. C. Batini, C. Cappiello, C. Francalanci, and A. Maurino, ACM Computing Sur-

veys CSUR 41, 16 (2009).3. A. Even and G. Shankaranarayanan, Journal of Computer Information Sys-

tems 50 (2009).4. M. H. Ofner, B. Otto, and H. Österle, Business Process Management Journal

18, 1036 (2012).5. A. Haug, F. Zachariassen, and D. V. Liempd, Journal of Industrial Engineering

and Management 4, 168 (2011).6. P. Glowalla, P. Balazy, D. Basten, and A. Sunyaev, Process-driven data quality

management—An application of the combined conceptual life cycle model,Hawaii Internasional Conference on System Science (2014), Vol. 47, p. 9.

7. P. Glowalla and A. Sunyaev, Managing Data Quality with ERP Systems-Insights from the Insurance Sector (2013).

8. Y. W. Lee and D. M. Strong, Journal of Management Information Systems20, 13 (2003).

9. T. C. Redman, Data Quality Management Past, Present, and Future: Towardsa Management System for Data: Handbook of Data Quality Research andPractice, Springer, Berlin, Heidelberg (2013).

10. G. Taguchi, Introduction to Quality Engineering: Designing quality into Prod-ucts and Processes (1986).

11. J. M. Juran, Juran on Leadership for Quality, Simon and Schuster (2003).12. T.-V. Tran, S. Kim, and M.-H. Hsiao, Data Stewardship and Flow Management

for Data Quality Improvement.

3063


13. M. F. Bosu and S. G. MacDonell, A taxonomy of data quality challenges inempirical software engineering, 22nd Australian Software Engineering Con-ference ASWEC, 2013 (2013), pp. 97–106.

14. L. Sebastian-Coleman, Measuring Data Quality For Ongoing Improvement:A Data Quality Assessment Framework, Newnes (2012).

15. K. S. Ryu, J. S. Park, and J. H. Park, ETRI Journal 28, 191 (2006).16. T. C. Redman, Digital Press (2001).17. R. Y. Wang and D. M. Strong, Journal of Management Information Systems

5, (1996).18. Y. Wand and R. Y. Wang, Communications of the ACM 39, 86 (1996).19. L. L. Pipino, Y. W. Lee, and R. Y. Wang, Communications of the ACM 45, 211

(2002).20. T. C. Redman and A. Blanton, Artech House Inc. (1997).21. G. Shankaranarayan, M. Ziad, and R. Y. Wang, Journal of Database Manage-

ment 14, 14 (2003).22. J. Herbsleb, D. Zubrow, D. Goldenson, W. Hayes, and M. Paulk, Communica-

tions of the ACM 40, 30 (1997).23. J. Kaur, Comparative Study of Capability Maturity Model (2014).24. H. U. Buhl, M. Röglinger, S. Stöckl, and K. S. Braunwarth, Business and

Information Systems Engineering 3, 163 (2011).25. J. V. Brocke and M. Rosemann, Handbook on Business Process

Management 2, Springer (2010).

26. L. P. English, Improving Data Warehouse and Business Information Quality,John Wiley & Sons (1999).

27. S. Muthu, L. Whitman, and S. H. Cheraghi, Business process reengineer-ing: A consolidated methodology, Proceedings of the 4th Annual InternationalConference on Industrial Engineering Theory, Applications, and Practice,1999, US Department of the Interior-Enterprise Architecture (2006).

28. A. V. Looy, M. D. Backer, and G. Poels, Enterprise Information Systems 8, 188(2014).

29. S. Bagchi, X. Bai, and J. Kalagnanam, Data Quality Management Using Busi-ness Process Modeling, Google Patents (2006).

30. W. M. V. D. Aalst, A. H. T. Hofstede, and M. Weske, Business processmanagement: A survey, Business Process Management, Springer (2003),pp. 1–12.

31. W. Bandara, D. R. Chand, A. M. Chircu, S. Hintringer, D. Karagiannis, andJ. C. Recker, Communications of the Association for Information Systems27, 743 (2010).

32. R. K. Ko, S. S. Lee, and E. W. Lee, Business Process Management Journal15, 744 (2009).

33. K. M. Hüner, M. Ofner, and B. Otto, Towards a maturity model for corpo-rate data quality management, Proceedings of the 2009 ACM Symposium onApplied Computing (2009), pp. 231–238.

34. A. Even and G. Shankaranarayanan, Journal of Data and Information QualityJDIQ 1, 15 (2009).


3064




Vol. 21, 3070–3074, 2015

Context-Aware Mobile Learning Model for Traveler

Dadang Syarif Sihabudin Sahid1�2�∗, Lukito Edi Nugroho1, Ridi Ferdiana1, and Paulus Insap Santosa1

1Department of Electrical Engineering and Information Technology, Universitas Gadjah Mada, Yogyakarta, 55281, Indonesia2Department of Computer, Politeknik Caltex Riau, Pekanbaru, 28265, Indonesia

Currently, the emerging of ubiquitous computing has provided a significant opportunity for using a mobile tech-nology as a learning medium. This situation has led many researchers to investigate the potential of adaptivelearning through mobile learning. However, there are limited researches present a model of adaptive mobilelearning for learners who are always traveling. This paper proposed a context-aware mobile learning modelfor learners who traveling frequently by considering several challenges in traveling situation. This model aimsto provide alternative learning by providing learning resources as well as the location presenting the nearestlearning resources tailored to conditions and situations context of learners. Learning style (video, audio, text,animation), learning preference (case-study, conceptual, simulation) and preferred time are selected as staticcontext parameter and used as an adaptive filter to presenting learning materials through mobile. On the otherhand, location and environment level (connectivity, noise, illumination) are chosen as the dynamic context toprovide the position of the nearest learning resources such as libraries and classmates for collaboration. As theresult of this model provided a wireframe design of context aware mobile learning for the traveler, by providingand recommending the learning materials, learning approach, as well as a nearest source of learning accord-ing to the conditions and situations of learners. In conclusion, this model can provide alternative guidelines fordevelopers who interested implementing context aware mobile learning, especially for frequent traveling learnerto keep earning and following the learning process wherever located.

Keywords: Mobile Learning, Context Aware, Context of Learning, Context of Mobile, Traveling Learner.

1. INTRODUCTIONEnhancing and delivering learning paradigm has been shifted

from traditional all-size type of learner to adaptive and personal-

ized learning. The last paradigm tailors individual learner’s needs

that are fitted with situation, preferences and context of learners.1

These needs include flexibility and mobility for the learner.

In addition, increasing capability the devices and data commu-

nication providers offered opportunities to access the internet

easily. This situation led researchers continuously investigating

and developing the learning activities via mobile device.

Mobile learning enacts feasible correlation between the mobile

device technology and the process of education. As mobile

devices and other portable transmission devices extent persis-

tently, the real worth of mobile learning can be examined in the

implementation of educational.2

Currently, using mobile in the learning process (mobile learn-

ing) gives advantages for the learners, especially for who has

mobile activities. The learners can adapt and personalize the

learning activities due to traveling schedule, place and per-

sonal situation. Many researchers have been experimented on

adaptation and personalization in mobile learning. The research

focused on how the context aware technology supports adaptive


and personalized mobile learning in ubiquitous and pervasive

computing.3�4 Context aware mobile learning systems capable

for providing the most suitable for presenting recommendations

and solutions to learners based on their learning styles and

preferences.

Adaptive and personalized mobile learning has been impor-

tant and desired complex research as well as issues in practices.

It should be considered that without proper functional model

design, integrating contextual information, adaptation and per-

sonalization process as well as presenting adaptation results as

appropriate recommendations for the learners can be compli-

cated and hard to be carried out. Moreover, the complexities

of the needs and preferences, situation, learning environment,

processes, included implementation methods increases drastically

when taking into account the support of mobile learners with

various of mobile devices accessing adaptive and personalized

recommendations. However, existing researches on context aware

mobile learning for traveler are still limited due to some issues

and challenges such as in designing, processing, communicating,

presenting previous e-learners to mobile learners, and traveling

situation contexts.5

This paper presented a context aware mobile learning model

for the traveler by describing the design of contextual learner,

adaptation process and presenting adaptation results. The model



also considers challenges related to traveling situation contexts,

such as connectivity, level of noise and level of illumination.

The following sections will be arranged from mobile learning

concepts, related work of context aware mobile learning, model

design of mobile learning, the proposed contextual model and

wireframe design of the proposed model.

2. THE CONCEPT OF MOBILE LEARNINGThe wide spread of various mobile devices and increasing wire-

less technology, the learners are shifted from previous e-learning

to the new learning experience through mobile education, namely

mobile learning. As the simple definition, mobile learning is an

e-learning via portable or moveable devices such as PDA, smart

phone, and tablets. Comparing with other form of distance learn-

ing components, the position of mobile learning can be shown

in Figure 1.6 Generally, distance learning give flexibility to the

learners who could not find a suitable place and time to attend

in the regular classroom due to far away from the place or work

activities. The distance learning offered learning opportunities

through various types of learning forms that can be classified into

e-learning and blended learning.

Mobile learning is a cutting-edge kind of learning process, uti-

lizing mobile devices, reaching various digital learning resources

and services, whenever, wherever. Mobile learning picks up and

conducts a learning environment that can be assisted by the

wireless technologies.7 The mobile learning purposed to occupy

learners in learning process stimulated by individual situation

and condition.8 Compared with the traditional learning methods,

mobile learning has following features: mobility, real-time, inter-

active, virtualization, digitization, and personalization.9 Different

with online learning, the mobile learning usually uses wireless

communication technology, easy to move, easy to carry.10 How-

ever, it is influenced by level of noise, quality of network and

level of illumination.

3. RELATED WORKS OF CONTEXT

AWARE MOBILE LEARNING AND

MOBILE TRAVELINGEmbedding context aware technology in mobile learning,

enabling capability of mobile learning can be adaptive and per-

sonalized. It is related to mechanism in the system that to reach

Fig. 1. Components of distance learning.6

the learning needs and preferences such as learning objectives,

previous knowledge of learners, learning styles, and situation

of learning environment such as current spot, weather, time, or

movements of interconnected learners.11 Term of adaptive in con-

text aware accords with the catching situation of learners, learn-

ing needs, and particular condition for considering to generate

appropriate model learning adaptations. Term of personalization

is generally related to customizing of the system capabilities,

incorporating also the learners concerns that can be pointed out

and adapted. The personalization included interface of the sys-

tem, expected language, or other concerns that cause the system

to become more personal.12

Some researches and developments related to context aware

adaptive and personalized mobile learning have been carried out

and given new learning experience for the learners. Yau et al.13

investigated learning preferences that suitable with the learners

needs and interests as a foundation for building a context aware

personalized mobile learning application based on mobile learn-

ing preferences. The research presented three kinds of learning

preferences included location of study, time of the day to study,

and the level of noise/distraction. Wang et al.14 applied context

aware technology and recommendation algorithm to personalize

learning resources and materials. Ako-nai et al.15 proposed the

5-R framework for mobile learning adaptation. The 5-R deals

and focuses on the 5-Rights: location, device, time, learner, and

contents. Zervas et al.16 presented a context-aware mobile learn-

ing to transform the learning resources that can be personalized

into a mobile player platform.

Context aware mobile traveling, mostly have been imple-

mented in the tourism recommender system instead of mobile

learning for traveling learners. Nevertheless, they have given the

inspirations and solutions if be implemented for building context

aware mobile learning for traveling learners. Yu et al.17 applied a

recommender system for traveler by providing a personal travel-

ing plan via a context-aware mobile technology. Gavalas et al.18

proposed a systematic approach for mobile recommender sys-

tem in tourism that is tailored the user needs. The recommender

systems included attractions, tourist services, collaborative user-

generated content and social networking, routes and tours, and

personalized multi-day tour planning. Borras et al.19 implemented

a survey for the intelligent tourism recommender system. The

survey provided a break down and current area that dealing with

various interface types, the widespread algorithms, capabilities

of the systems and their use of advanced methods.

4. DESIGN OF CONTENT ADAPTATION FOR

CONTEXT AWARE MOBILE LEARNINGThe design adaptive and personalized context aware mobile

learning is a mechanism to convey learner contextual information

and kind of recommendations that can be produced accordingly

as shown in Figure 2. It led to investigate more detail about what

the key components in designing context aware mobile learning

application. The components are learner mobile context that con-

sisted learner mobile contextual information, adaptation engine

and adaptation personalized of mobile learning.20–22

4.1. Learner Contextual Information for

Mobile Learning

In terms of general e-learning, the contexts consist of static con-

text and dynamic context, whereas in mobile learning, learner

3071


Fig. 2. Mechanism of adaptive and personalized context aware mobile learning.22

contextual information become more specific and be classified

into context of learning and context of mobile. Generally, the

context of learning and the static context related to the con-

texts that are input by the learner and more constant, while the

dynamic context and the mobile context, mostly related to the

sensing capability of the system or devices. Static context as

shown in Table I can be segregated into three sub-contexts:23

personal, abstraction, and situation context.

Another context model combined context of learning and con-

text of mobile to describe learner contextual information. The

main dimensions and elements of the contexts can be shown in

Table II. Context of learning is related to a certain design of

learning and profile of individual learner, while the context of

mobile is related to context of learning that captured and took by

sensor in mobile devices.24

4.2. The Engine of Context Aware Mobile Learning

The engine of context aware mobile learning similar to process-

ing unit that acquires input data from learner contextual infor-

mation and produces the educational content adaptation results.

There are two main adaptation engines with several approaches

that can be implemented for context aware mobile learning

systems:22

• Adaptation rules, that is using flow statements and conditional

form (if-then, if-then-else statements) generated by certain con-

text of learning or context of mobile.

• Adaptation algorithms, that is using various kinds of algo-

rithms such as intelligent approaches, heuristic algorithm,

Table I. Static context aware parameter.

Static context Parameter

Personal contextPersonal information Name, ID, Date of birth, address, gender,

email, phone number, technologiesknown, knowledge level, OSexperience, internet usage

Personality type Extrovert, sensory, thinkers, judgersLevel of expertise Beginner, practitioner, expert

Abstraction contextLearner preference Conceptual, example-oriented, case-study,

simulation, demonstrationLearner intention Research, survey or overview, quick

reference, basic introduction, project,assignment, seminar

Learning style Video, audio, text, animation, slidesSituation contextLearner situation Private, public, drivingNetwork Wired, wirelessDevice Mobile, PDA, laptop, PCQuality of learning service Functional requirements, non functional

requirements

decision-based algorithms, matching algorithm, artificial algo-

rithm, and similarity algorithm, to proceed particular context of

learning or context of mobile.

4.3. Adaptation of Context Aware Mobile Learning

There are two primary classifications of adaptation in context

aware mobile learning: (1) adaptation related to resources of

learning; (2) adaptation related to activities of learning.25 Each of

classifications is divided into sub-category or kind of adaptations.

The first classification consisted of:

• Selecting, it is carried out by filtering relevant materials of

learning and provided to the learners based on context of learning

and context of mobile.

• Presenting, it is how to show the learning material through

mobile devices.

• Navigating and sequencing, it is rearranging or reordering

resources to make the individual learning pathways.

The second classification of adaptation context aware mobile

learning consisted of:

• General adaptation, it is generating activities of per-

sonal learning automatically, according to several require-

ments captured from context of learning and context of

mobile.

• Feedback and support, it is an adaptation by providing individ-

ual advises and suggestions related to learning time and learning

activities.

Table II. Dimension elements of context.24

Dimensions Contexts

Context of learningDesign of learning Objectives of learning, pedagogical approaches,

activities of learning, roles of participation,tools, and resources of learning.

Profile of learner Expertise (skills, attitudes, knowledge), role,individual condition and characteristics(learning needs, learning styles, learninginterests, physical condition or other inabilities).

Context of mobileLearner Temporal and situational condition (preferences,

mood, needs, interests).People Role, relationship, contributions and constraints.Place Location and position, zones, room of interactive,

setting and culture of learning.Artifact Technological and non-technological aspect.Time Duration of task, schedule of task, availability,

and action happens.Environment Network quality, illumination, noise level, weather

conditions.

3072


• Navigation to locations, it is and adaptation by providing

awareness of place and position related to appropriate activities

of learning.

• Communication and interaction, it is an adaptation by facili-

tating the learners while executing activities of learning.

5. THE PROPOSED MODEL OF CONTEXT

AWARE MOBILE LEARNING FOR

TRAVELERPrevious concept, related works, tools and design model for con-

text aware adaptive and personalized mobile learning has been

given suitable and proper foundation to be implemented in con-

text aware mobile learning for traveling learner. The concept

of mobile learning gives basic knowledge and characteristics of

mobile learning. Related works in mobile learning application

shows the state of the art about researches in adaptive and per-

sonalized context aware mobile learning, whereas recommender

mobile applications in tourism gives inspirations, how the con-

text aware mobile learning implemented for travelers especially

how to deal with the constraints of traveling situation context

such as limited connectivity, noise and illumination level. Design

adaptation of context aware mobile learning gives guidance and

become tools to develop the proposed model. Figure 3 shows the

components and elements for the proposed context aware mobile

learning model for traveler.

In the learner contextual information, the proposed model

using learning styles, learning preferences, and time preference

as learning context that is input by the learner. Location with

GPS technology sensor is used as mobile context to detect near-

est library and classmates location. As well as the Wi-Fi, noise

and illumination sensor will be used by the system to select

appropriate learning materials. In adaptation engine phase, adap-

tation rules such as condition structure rule based are imple-

mented as the approach. In this phase, also are provided with

library database, learner profiles, map services, GPS services

and resources database. In the adaptation phase, some adapta-

tion types are conducted. Selection adaptation type is used for

Fig. 3. Components and elements of the proposed model.

Fig. 4. Wireframe design of the proposed model.

presenting learning material recommendations based on learning

style and learning preferences by dealing with the connectivity

level (poor, good, strength), noise level, and illumination level.

On the other hand, navigation to location, communication and

interaction adaptation type are used for presenting library and

nearest classmates location.

3073


In order to give more description about the proposed model,

this paper also provides simple wire frame design as can be

shown in Figure 4. The wire frame design is a low fidelity pro-

totype design to figure out the model if will be implemented.

6. CONCLUSIONThe proposed model of context aware mobile learning for the

traveler has been presented in this paper. The model has been

developed based on adaptive design as a tool of adaptive and

personalized context aware mobile learning. The tool has given

suitable knowledge for building a context aware mobile learning.

Traveling learner is chosen as a model implementation subject

due to limitation of researches and works in this area due to

some issues and challenges related to traveling situation contexts.

All processes in building the proposed model has been carried

out as a foundation for implementation. The proposed model is

expected not only improving the previous context aware mobile

learning model, but also contributing to learners especially who

have traveling activities frequently. In the future works, the model

can be implemented in a real situation by considering more com-

plex of the contexts, adaptation engines, and type of adaptations.

References and Notes1. S. Gómez, P. Zervas, D. G. Sampson, and R. Fabregat, Delivering adaptive

and context-aware educational scenarios via mobile devices (2012).2. Y. Jiugen, X. Ruonan, and W. Jianmin, Applying research of mobile learning

mode in teaching 417 (2010).3. K. Chin and Y. Chen, The 2nd International Conference on Integrated Informa-

tion a Mobile Learning Support System for Ubiquitous Learning Environments,Procedia—Soc. Behav. Sci. (2013), Vol. 73, pp. 14–21.

4. A. Syvänen, R. Beale, M. Sharples, M. Ahonen, and P. Lonsdale, Support-ing pervasive learning environments: Adaptability and context awareness inmobile learning 5 (2005).

5. Y. Chia and F. S. Tsai, Context-aware mobile learning with a semantic service-oriented infrastructure 896 (2011).

6. S. Karadeniz, Flexible design for the future of distance learning, Procedia—Soc., Behav., Sci. (2009), Vol. 1, pp. 358–363.

7. I. Conference and D. Learning, M-learning: An innovative advancement of ICTin education 148 (2010).

8. S. Chin, Mobile technology and gamification: The future is now! 138(2014).

9. Y. Jin, Research of one mobile learning system 162 (2009).10. S. Behera, International Journal on New Trends in Education and Their Impli-

cations 65 (2013).11. S. Wu, A. Chang, M. Chang, T. Liu, and J. Heh, Identifying personalized

context-aware knowledge structure for individual user in ubiquitous learningenvironment 95 (2008).

12. R. M. Carro, Supporting the development of mobile adaptive learning envi-ronments: A case study 2, 23 (2009).

13. J. Y. Yau and M. Joy, A context-aware personalized m-learning applicationbased on m-learning preferences.

14. S.-L. Wang and C.-Y. Wu, Expert Syst. Appl. 38, 10831 (2011).15. F. Ako-nai, Q. Tan, and F. C. Pivot, The 5R adaptive learning content gener-

ation platform for mobile learning (2012).16. P. Zervas and D. G. Sampson, Context-aware adaptive and personalized

mobile learning delivery supported by UoLmP 47 (2014).17. C.-C. Yu and H. Chang, Towards Context-aware recommendation for per-

sonalized mobile travel planning, Context-Aware Systems and ApplicationsSE-12, edited by P. Vinh, N. Hung, N. Tung, and J. Suzuki, Springer, Berlin,Heidelberg (2013), Vol. 109, pp. 121–130.

18. D. Gavalas, C. Konstantopoulos, and K. Mastakas, J. Netw. Comput. Appl.39, 319 (2014).

19. J. Borràs, A. Moreno, and A. Valls, Expert systems with applications intelligenttourism recommender systems: A survey 41, 7370 (2014).

20. M. F. Fudzee, A classification for content adaptation system 426 (2008).21. J.-M. Su, S.-S. Tseng, H.-Y. Lin, and C.-H. Chen, User Model. User-Adapt.

Interact. 21, 5 (2011).22. P. Zervas, D. Sampson, S. Gómez, and R. Fabregat, Designing tools

for context-aware mobile educational content adaptation, Ubiquitous andMobile Learning in the Digital Age SE-3, edited by D. G. Sampson,P. Isaias, D. Ifenthaler, and J. M. Spector, Springer, New York (2013),pp. 37–50.

23. M. M. Das, Static context model for context aware E-learning 2, 2337 (2010).24. P. Zervas, S. Eduardo, G. Ardila, R. Fabregat, and D. G. Sampson, Tools for

context-aware learning design and mobile delivery 3 (2011).25. D. Sampson and P. Zervas, Context-aware adaptive and personalized mobile

learning systems, Ubiquitous and Mobile Learning in the Digital Age SE-1,edited by D. G. Sampson, P. Isaias, D. Ifenthaler, and J. M. Spector, Springer,New York (2013), pp. 3–17.


3074




Vol. 21, 3080–3084, 2015

Prediction of CO2 Emissions Using an Artificial

Neural Network: The Case of the Sugar Industry

Chairul Saleh1�∗, Raden Achmad Chairdino Leuveano2, Mohd Nizam Ab Rahman2,Baba Md Deros2, and Nur Rachman Dzakiyullah3

1Department of Industrial Engineering, Faculty of Industrial Technology, Universitas Islam Indonesia,Yogyakarta, 55584, Indonesia

2Department of Mechanical and Material Engineering, Faculty of Engineering and Build Environment,Univeriti Kebangsaan Malaysia, Malaysia

3Department of Informatics Engineering, Faculty of Engineering, Janabadra University, Yogyakarta, Indonesia

In this paper, the back-propagation artificial neural networks (ANN) model is presented to predict expenditureof carbon (CO2) emission. The model was built based on the input variables that affect to expenditure of CO2

include the amount of bagasse, wood and marine fuel oil used in boiler machine. The objective of this paperis to monitor the CO2 emission based on the fuel used for operating the boiler machine. The data used fortesting the models were obtained from Sugar Industry. It splits up into 90% of training data and 10% of testingdata. The model experiment was conducted using trial and error approach to find the optimal parameters ofANN model. The result shows that the architecture of ANN model have optimal parameter on training cycle 50,learning rate 0.1, momentum 0.1, and 19 hidden nodes. The validity of the trained ANN is evaluated by usingRoot Mean Square Error (RMSE) with error value as 0.055. It indicates that the smallest error provides moreaccurate results on prediction and even can contribute to the industrial practice, especially helping the executivemanager to make an effective decision for business operation by considering the expenditure of CO2 monitoring.

Keywords: Artificial Neural Network, Prediction, Carbon Emission, RMSE.

1. INTRODUCTIONThe global warming effect poses a significant threat to the envi-

ronment. According to Research Working Group III of the Inter-

governmental Panel on Climate Change (IPPC), a variety of

living creatures, including humans, are the highest producers of

greenhouse gases causing global warming.1 Greenhouse gases

such as carbon dioxide, methane and nitrous oxide act as a

blanket layer in the Earth’s atmosphere, trapping the sun’s heat

and increasing the temperature of the Earth. As stated in the

fourth assessment report from the IPCC,2 fossil fuels used by

human activities in the energy sector, transportation, agriculture

and industry have been linked to the increase in greenhouse

gases.

As a key sector of production and the backbone of the

economy in Indonesia, the manufacturing industry contributed

118.12 MtCO2 in 2011.3 From these data, it is apparent that large

volumes of energy are consumed to produce goods or services

within the economy, but also wastes are generated from such

production operations. As a result, there will be a continuous

increase in CO2 emissions, categorized as air waste/pollutants.


Therefore, the Indonesian government needs to address this by

controlling industries’ emissions.

As one of the large contributors to CO2 emissions, the sugar

industry is here selected as a case study. In 2008, 3.92 million

tonnes of sugar was produced in Indonesia.4 The more sugar pro-

duced, the more CO2 emissions are potentially generated. These

emissions can come from the fuel used in boilers and the con-

sumption of electricity, solar energy and liquid petroleum gas

(LPG). For instance, boilers are fired by fuel to generate steam.

This steam then used to produce electricity using a generator to

change the energy from heat to mechanical energy. The electric-

ity produced by the boiler is then used to operate the cane cutter

and hammer shredder to crush the sugar cane in the first stage

of the process. The combustion of fuel in the boiler emits CO2,

potentially generating air pollutants. This relationship between

polluting emissions and energy consumption has been studied by

Pao and Tsai.5

It is necessary to monitor machines in industrial operations

as each machine can contribute different amount of emissions.

As boilers depend on fuel combustion, potentially high in emis-

sions, this study focuses on boilers and investigates the variables

that affect CO2 emissions from these machines. The three types

of fuel used to operate boilers include bagasse (fibrous sugarcane



residue), firewood and marine fuel oil (MFO). Monitoring these

fuels is very important because they have different effects on the

emission of CO2.

To monitor machine emissions, most studies use regression,

discriminant analysis and artificial neural networks (ANN) in

empirical modelling to control the system.6–8 Of these, the ANN

model has proved the most attractive to researchers for pre-

dicting the behaviour of a system in certain cases and neural

networks have increasing been used in recent years over conven-

tional statistical techniques such as regression and discriminant

analysis.9 The model includes machine performance testing, cut-

ting mechanics, signal processing, data decomposition and image

processing.10 The ANN model is a statistical learning algorithm,

the design of which was inspired by the properties of biologi-

cal neural systems, used to search and produce new knowledge

to estimate functions based on a large number of inputs.11 The

ANN model can also be defined as a mathematical model for

predicting new problems.12

As harmful emissions have increased, previous research has

also attempted to embed the use of ANNs in environmental appli-

cations. Baareh13 applied the ANN model to estimate CO2 emis-

sions for consumption from four fuel inputs, including global oil,

natural gas (NG), coal and primary energy (PE). Fontes et al.14

employed ANN model to classify ozone episodes, which have a

negative impact on the environment. This aimed to reduce ineffi-

ciency leading to ozone precursor emissions. To prevent fouling

problems in machines resulting in higher CO2 emissions, Romeo

and Gareta15 and Rusinowski and Stanek16 used the ANN method

to monitor boiler performance. As the ANN model is a powerful

tool for handling such types of modelling processes, this paper

proposes an ANN model to predict CO2 emissions in boilers in

the sugar industry. The objective is to monitor CO2 emissions

based on fuel combustion used to operate the boilers in sugar

production. By monitoring the machines, the sugar industry can

efficiently manage fuel combustion.

The remainder of this paper is organized as follows: Section 2

introduces the materials and methods in the design of the ANN

model; Section 3 presents the result of the experiment; finally,

Section 4 summarizes the salient points and concludes the paper.

2. MATERIALS AND METHODSTo design the ANN model for predicting the CO2 emissions from

boiler use, a number of steps were defined as shown in Figure 1.

The first step to in the design of the ANN model is data col-

lection. The primary data were collected from the sugar indus-

try in Yogyakarta. This research analyses boiler emissions based

on the fuel used. The three main fuels that affect CO2 emis-

sions are found to include bagasse, firewood and MFO. These

fuels are thus categorized as input variables. The output vari-

able (CO2 emission) is computed by multiplying activity data

(e.g., fuel consumed) by the emission factor for that activity,

in accordance with the guidelines for computation of emissions

provided by the IPCC.17 The relationship between the input and

output variables is defined as:

Xi =

∣∣∣∣∣∣∣Bagasse�X1

Firewood�X2

MFO�X3

∣∣∣∣∣∣∣ � Y = CO2 (1)

Dataset cleansing, transformation-normalization

Cross-validation

Training Set Test Set

Data collection

• Bagasse• Firewood• MFO• Carbon emission

12

10....................................

Total Data

k-fold cross-validation

Model Search

Modelevaluation

Fig. 1. Design overview of the ANN model.

To avoid any noisy data, missing data, incorrect, improp-

erly formatted, or duplicated in datasets, then data cleansing is

employed.22 The aim of data cleansing is to have a better repre-

sentative datasets for developing reliable neural network model

and improve the accuracy of prediction. The process of data

cleansing in this study was automatically performed by Rapid

Miner software. However, as shown in Figure 2, the cleansing

process shows that dataset has zero noisy and missing value. The

next step is to transform and normalize the dataset in order to

have inputs with 0 means and a standard deviation of 1.

Then, to validate the model, k-fold cross-validation is used.

This technique divides the dataset into a training set and a test

set. The training set is used to calculate the gradient and update

the network weights and biases. In this case, validation was com-

puted during the training process to obtain minimum error in

prediction. The test dataset is used to assess the performance of

Fig. 2. The process of data cleansing using rapid miner.

3081


a fully-trained ANN model with data outside the training set.

To find the optimal architecture of the ANN model in the training

process, the parameters—training cycle, learning rate, momen-

tum and hidden node—must be optimal. The simplified architec-

ture of the ANN model comprising three layers (input variables,

hidden nodes and output variable) is shown in Figure 3.

A back-propagation (BP) learning algorithm is then used in the

training process to obtain the optimal parameters. BP training is

categorized as a gradient descent algorithm.7 With this technique,

the total error can be reduced by changing the weight along its

gradient. However, to analyse the effect of network parameter,

a trial and error approach is used in the design of the BP neural

network prediction by varying the network structure based on the

pre-processing of the data, the number of input nodes and the

activation function. To evaluate the model error, the root mean

square error (RMSE) is expressed by:

RMSE = 1

2

[∑j

∑k

�Yjk −Ojk�]1/2

(2)

where Y is the predicted value and O denotes the actual value

vectors over pattern k.

3. RESULTSRapid Miner 5.2.003 was to carry out the experiment. All the test

problems were undertaken using a computer with the following

specifications: Intel (R) Core (TM) i5-2450M CPU @ 2.50 GHz

(4 CPUs), 8 GB of RAM and Windows 8.1, 64-bit (6.3, Build

9600) operating system. As mentioned previously, the data for the

experiment were obtained from the sugar industry in Yogyakarta,

Indonesia. There are 124 datasets for the period 2009–2013 on

the use of fuels, including bagasse (tonnes), firewood (tonnes)

Fig. 3. Simplified architecture of the ANN model, adopted from Erchan andAtici.20

Table I. Optimal parameters of the back-propagation ANN model.

Parameters Value

Training cycle 50Learning rate 0�1Momentum 0�1Hidden nodes 19RMSE 0�055

and MFO (litres), for boiler operation. From the data on fuel

consumed, the next step is to calculate the CO2 emissions (tonnes

CO2) using the IPCC17 (2006) guidelines. Our objective is to

monitor CO2 emissions from boilers, seek an accurate prediction

that has the lowest error. In other words, an accurate prediction

can provide information regarding the fuel consumption that has

the lowest emissions.

In this experiment, 10-fold cross-validation was used to split

the data into a training set and a test set. The data were divided

into 10 sets of size n/10 or equal parts: 90% of the data were

used for the training set and 10% for the testing set. Error evalu-

ation was then performed on 90/10 splits, repeated for all 10 pos-

sible splits. For each repetition, the training fold was normalized.

The mean values and standard deviations were taken over k dif-

ferent partitions. The k-fold cross-validation technique was used

as it can provide a lower variance, meaning that minimum error

can be achieved.

As noted above, the ANN model consists of three layers, for

which the parameters include the training cycle, the learning rate,

momentum and hidden nodes. Using the BP algorithm with a trial

and error approach for the training process, the optimal architec-

ture of the ANN model is as shown in Table I.

The objective of the training process is to minimize the error

of prediction. This is taken as a rule to choose the optimal param-

eters of the ANN model. As can be seen from Table I, the perfor-

mance of the ANN model has an error value (RMSE) of 0.055.

The closer the error value is to zero, the higher the accuracy of

prediction. When the error value achieved its minimum, the train-

ing cycle was terminated at 50 cycles. The learning rate of the

ANN model was 0.1. The value of the learning rate represents

0.000

0.200

0.400

0.600

0.800

1.000

1.200

1 9 17 25 33 41 49 57 65 73 81 89 97 105

113

121

Err

or V

alue

Data

Actual

Prediction

Fig. 4. Comparison of actual values versus predicted values.

3082


Table II. Test of normality.

Shapiro–Wilk

Statistic df Sig.

0.977 124 0.032

the speed at which the system learns (converges). The momen-

tum rate has the same value as the learning rate, which is 0.1.

The momentum rate is used to avoid local minima and significant

changes in weights so that the global minimum can be achieved.

The optimal hidden nodes in the network numbered 19. Deter-

mining the optimal number of hidden nodes can avoid overfitting

and underfitting of the model. Overfitting occurs when the learn-

ing algorithm captures noise from the data, whereas underfitting

occurs when the learning algorithm cannot correctly detect the

underlying trend of the data, so that the model shows low bias but

high variance. Based on the results of the experiment, Figure 4

shows the actual and predicted values for both training and test

cases.

In the last step, the model output of the neural network is

analysed. This analysis aims to determine the level of confidence

in the trained model by seeing whether the predicted values are

normally distributed.18 If the predicted value is normally dis-

tributed, the model prediction has higher accuracy and precision.

Several statistical tests can be employed to check the normal-

ity of distribution, including the Shapiro–Wilk test, the Lilliefors

test, D’Agostino–Pearson’s L2 test, the Jarque–Bera test, and the

Anderson–Darling test. However, the most applicable statistical

method that fits all types of distribution and sample size is the

Shapiro–Wilk test19 and thus it was used in this study to test nor-

mality. In this paper, the significance level (�) was set at 0.05; if

p < 0�05, the data are not normally distributed. The result of the

Shapiro–Wilk test is shown in Table II.

As can be seen from Table II, the significance is 0.032. Thus,

the predicted values are not normally distributed. However, if

the statistical test does not show normal distribution, histogram

analysis can be used, as shown in Figure 5.

Basically, statistical data analysis is used to recognize the sta-

tistical probability distribution of the data. As the result, statisti-

cal inference and information based on data can be obtained and

Fig. 5. Histogram of normality distribution.

help to make the right decision.21 From Figure 5, it transpires

that despite the significance value found above, the prediction

results approximate normal distribution. Although, the prediction

model has higher accuracy that shown by lower RMSE value,

however, the histogram is centred over the true value. The result

shows the prediction model has poor repeatability and poor pre-

cision. However, the output of the ANN model is still proven to

solve non-linear data and provide higher prediction accuracy.

4. CONCLUSION AND RECOMMENDATIONThe back-propagation ANN model was applied in this paper to

predict CO2 emissions from boiler operations. The prediction

model is used to monitor fuel combustion from bagasse, fire-

wood and MFO, which affect the amount of CO2 emitted. The

ANN model was designed with three layers (input variable, hid-

den nodes and output variable). To obtain better prediction with

a lower error (RMSE) value, the trial and error approach was

applied. The minimum error value can also be used to optimize

the parameters of the ANN model. The results obtained show

that the RMSE value was 0.055 with optimal parameters for the

ANN model of 50 for the training cycle, 0.1 for the learning

rate, 0.1 for the momentum rate and 19 for the number of hidden

notes.

Greater accuracy in prediction can provide accurate informa-

tion regarding CO2 emissions. It means that when creating the

model prediction, the main goal is to achieve the lower RMSE

value. It can be conclude that lower RMSE value, greater accu-

racy of prediction can be obtained. As the result of this predic-

tion, this study can help manager to monitor boiler machines and

then develop policies or take decisions regarding production that

reduce the negative impact on the environment. By this reason-

ing, this study can contribute to practice in monitoring machine

emissions using ANN model prediction. In further research, this

ANN model could be integrated with optimization techniques,

such as genetic algorithms, particle swarm optimization and ant

colony optimization to improve the accuracy of prediction.

Acknowledgment: The authors would like to thank the

anonymous reviewers for their valuable comments. This research

was supported by This research is supported by Directorate of

Research and Community Service and Board Academic Devel-

opment, Universitas Islam Indonesia.

References and Notes1. IPCC, Climate change 2014: Mitigation of climate change [Online] Available

from: http://report.mitigation2014.org/report/ipcc_wg3_ar5_full.pdf [Accessedon 16th August 2014] (2014).

2. IPCC, Climate change 2007: The physical science basis [Online] Avail-able from: http://www.ipcc.ch/pdf/assessment-report/ar4/wg1/ar4-g1-spm.pdf[Accessed on 16th August 2014] (2007).

3. WRI, WRI’s climate data explorer: GHG emissions-energy-sub-sector [Online]Available from: http://cait2.wri.org/ [Accessed on 16th August 2014] (2011).

4. Industrial Department, Roadmap industri gula [Online] Available from:http://agro.kemenperin.go.id/e-klaster/file/roadmap/ [Accessed on 16th August2014] (2009).

5. H. T. Pao and C. M. Tsai, Energy 36, 2450 (2011).6. Y. Çay, A. Çiçek, F. Kara, and S. Sagiroglu, Applied Thermal Engineering

37, 217 (2012).7. S. A. Kalogirou, Progress in Energy and Combustion Science 29, 515 (2003).8. A. Maijidan and M. H. Zaidi, International Journal of Fatigue 29, 489

(2007).9. M. Paliwal and U. A. Kumar, Expert Systems with Applications 36, 2 (2009).

3083


10. A. Negarestani, S. Setayeshi, M. Ghannadi-Maragheh, and B. Akashe,Applied Radiation and Isotopes 58, 269 (2003).

11. H. Yoon, C. S. Park, J. S. Kim, and J. B. Baek, Expert Systems with Applica-tions 40, 231 (2013).

12. G. Zhang, B. E. Patuwo, and M. Y. Hu, International Journal of Forecasting14, 35 (1998).

13. A. K. Baareh, Journal of Software Engineering and Applications 6, 338 (2013).14. T. Fontes, L. M. Silva, M. P. Silva, N. Barros, and A. C. Carvalho, Science of

the Total Environment 488, 197 (2013).15. L. M. Romeo and R. Gareta, Applied Thermal Engineering 26, 1530 (2006).16. H. Rusinowski and W. Stanek, Energy Conversion and Management 48, 2802

(2007).

17. IPCC, IPCC guidelines for national greenhouse gas inventories. Energy (2)[Online] Available from: http://www.ipcc-nggip.iges.or.jp/public/2006gl/vol2.html [Accessed on 16th August 2014] (2006).

18. D. K. Williams, Jr., A. L. Kovach, D. C. Muddiman, and K. W. Hanck, AmericanSociety for Mass Spectrometry 20, 1301 (2009).

19. N. M. Razali and Y. B. Wah, Journal of Statistical Modelling and Analysis 2, 21(2011).

20. S. Ercan and U. Atici, Neural Computation and Application 22, 1039(2013).

21. C.-T. Su and C.-J. Chou, Quality Engineering 18, 293 (2006).22. C. Vercellis, Business Intelligence: Data Mining and Optimization for Decision

Making, Wiley Publisher (2009), ISBN: 978-0-470-51139-8.

Received: 19 November 2014. Accepted: 5 January 2015.

3084




Vol. 21, 3138–3142, 2015

Identification E-Learning Readiness in the Faculty

of Agricultural Technology Jambi University

Kurniabudi∗, SetiawanAssegaff, and Sharipuddin

STIKOM Dinamika Bangsa Jambi Jl. Jenderal Sudirman, Thehook, Jambi, Indonesia

Identification of E-learning readiness needs to be done so that the implementation of e-learning can workwell without spending money, time and effort. This study aims to identify the level of readiness of e-learningat the Faculty of Agricultural Technology Jambi University (abbreviated as FTP UNJA). Collecting data usingquestionnaires. Identify the level of e-learning readiness ELR model. The study concluded FTP UNJA be in aposition ready for the implementation of e-learning with some improvement in the perception of lecturers andmanagement policies.

Keywords: Identification, e-Learning, Readiness, ELR Models, Higher Education.

1. INTRODUCTIONAs one of the colleges in Jambi who realize the importance of the

implementation of ICT in supporting the learning process is the

Faculty of Agricultural Technology Jambi University (abbrevi-

ated as FTP UNJA). In the implementation of e-learning. Before

performing the implementation of e-learning organizations must

identify readiness of e-learning and planning a closely to prevent

failures such as excess costs or product is not attractive. In the

development strategy should be considered a matter of time, cost,

infrastructure and management.6�7

Therefore, the FTP UNJA need to develop a strategy so that

e-learning is developed in accordance with the objectives to be

achieved. One of the steps undertaken strategy is to identify

readiness E-learning in the Faculty of Agricultural Technology

UNJA. Identification was done by measuring the level of readi-

ness of e-learning at the FTP UNJA.

With identification of e-learning makes it possible to obtain

information level implementation of e-learning readiness and

identify factors that need to be considered in the implementation

of e-learning on the FTP UNJA that can be assembled design

e-learning strategy thoroughly and effectively implement IT des-

tination.

2. THEORY2.1. Overview of E-Learning

Today the development of Internet technology has ability to com-

municate multimedia platform that makes the development of

e-learning is becoming an effective mechanism for learning and


teaching. In education, e-learning offers a variety of benefits and

advantages for students and teachers in the activities of research,

training, and online learning. In many countries of e-learning

has become one of the components of the “importance” of the

lifelong learning strategy and long-term.3 In general, e-learning

is defined as learning method that uses electronic-based instruc-

tional content presented via the Internet.16 Another opinion states

that e-learning is an education and training delivered through

ICT, which is specifically designed to support the learning or

performance objectives of the organization.7

Basically, the use of electronic media and computer networks

in e-learning to improve access and quality of teaching and learn-

ing process. Based on the opinion of experts, then e-learning can

be interpreted as a learning system that utilizes information and

communication technology in order to reach the goal of effective

and efficient learning.

Based on some of the opinions can be concluded that

e-learning is a learning system where the process of interac-

tion between teachers as instructors and students using electronic

media. Electronic media in question is a stand alone PC or a

PC connected to a computer network can be either intranet and

internet. In e-learning materials electronically packaged, in the

form of a slide presentrasi or in the form of web pages that

usually combines several forms such as text, images, sound and

even video with material goals more easily understood and can

increase student interest. E-learning is a solution, which offers

ease of access, dissemination, and sharing of learning resources

with which to support various types of instruction.8

2.2. Assesment E-Learning Readiness

Prior to implementing e-learning program at an institution is

very important for the institution to carry out the measurement



of readiness for the implementation of e-learning.15 Clark and

Mayer provides guidance to managers of institutions that conduct

readiness assessment before adopting e-learning.3 A test (assess-

ment) readiness (readiness) is a measurement of the degree/level

of capability and capacity to a knowledge in a particular context.

Every group of people, or countries and different populations

have different ways to respond to the knowledge-based initiatives,

therefore readiness studies must take into account the impact that

give effect to the situation, condition or learning programs.11

2.3. E-Learning Readiness Studies

The topic of e-learning readiness is not a new topic. There have

been many studies conducted with regard to e-learning readiness.

Identification of e-learning readiness conducted to measure the

readiness of an institution in the implementation of e-learning as

presented in Table I.

The study of e-learning readiness is not only conducted in

educational institutions, but also can be done at the company,

group of individuals or communities which are part of the stake

holders in the e-learning system.

2.4. E-Learning Readiness Factors

In conducting the assessment of e-learning readiness to consider

aspects of the benchmark readiness. In previous studies in gen-

eral, any researcher using a similar aspect, but in different situ-

ations and organizations need to add other components adapted

to existing conditions. As with Ref. [5] identifies aspects of e-

learning readiness is divided into two levels. The first level is

a key component of the e-learning readiness is in government,

industry, education and social. The second level of readiness can

be measured by conectivity, capacity defined as the ability of a

Table I. E-learning readiness researches.

Authors, years Research topic Method Responden

Kaur and Abas (2004) An assessment of E-learningreadiness at open UniversityMalaysia

questionnaire (factors: management,personnel, content, technical,environtmental, cultural andfinancial)

Sample: tutor and learner fromdiploma, undergraduate andpost-graduate

Aydin and Tasci (2005) Measuring readiness for e-Learning:Reflection from an emergingcountry

Survey (factors: technology,innovation, people,self-development)

Director of the human resources inthe top 100 companies in turkey

Eslaminejad et al. (2010) Assesment of instrutors’ readiness forimplementing e-learning incontinuing medical education Iran

Questionnaire (factors: knowledge,attitude, skills and habits intechnology and pedagogy domain)

Teachers in medical academic

Saekow and Samson (2011) E-Learning readiness of Thailand’sUniversities comparing to theUSA’s cases

Survey (factors: policy, technology,financial, human resources,infrastructure)

Executive, deans, and technician fromhigher education in Thailand andUSA

Oketch (2013) E-learning readiness assessmentmodel in Kenyas’ higher educationinstitutions: A case study ofUniversity of Nairobi

Questionnaire (factors: technological,cultural, content)

Using sample: Lecture from theUniversity of Nairobi

Al-Furaydi (2013) Measuring e-learning readinessamong EFL teachers inintermediate Public Schools inSaudi Arabia

Survey (factors: Attitude towarde-learning, computer literacy)

Sample of the population: Teachers in24 schools in Al-Madinah

Azimi (2013) Readiness for implementation ofE-learning in colleges of education

Questionnaire (factors: ICTinfrastructure, human resources,budget and finance, Psychologicaland content)

Receiver and heads of college ofeducation affiliate by University ofMysore

Vikonis et al. (2013) Readiness of Adults to Learn usingE-learning, M-learning andT-learning technologies

Survey (factors: technological,attitudes towards the mode oflearning, experience in using othere-services)

Target group: employable (adults)residents of Lithuania

country to deliver and consume e-learning, literacy levels and

trends in education and training.

Rosenberg (2000b) in his research to identify aspects of

e-learning readiness is more focus on sustain ability (sustain

ability) and propose some aspects of business readiness such as

changes in the conventional learning and e-learning, the value

of the instruction and information, the role of change manage-

ment, the rediscovery of the training organization to support

effort e-learning, e-learning industry, and personal commitment,6

propose aspects of e-learning readiness include a component

of psychological, sociological, environmental, human resources,

financial, technology, equipment, and content readiness. Haney10

identify seven aspects of readiness on e-learning include: human

resources, learning management system, learners, content, infor-

mation technology, finance, and vendors. While Oketch,12 has

developed the model in Figure 1 Model developed a hybrid

model of the technological readiness,4�6 culture readiness,5�11 the

content readiness,5�6�13 and demographics factors.4

In this study, the authors adopted a model developed by Oketch

(2015) to identify the readiness of e-learning, because the author

considers aspects used more complete, because it has been incor-

porating aspects of e-learning readiness of some research on the

assessment readiness.

3. METHODOLOGYIn this study the authors used a model of e-learning readiness

assessment developed by Oketch. Oketch,13 with four main fac-

tors of e-learning readiness in the development of the ques-

tionnaires. Consisting of: Demographic, Tecnological Readiness,

Content and Culture Readiness Readiness. Oketch models used

because, instruments have been developed based on the research

3139


of e-learning readiness level as has been done by Chapnick

(2000), Borotis and Poulymenakou,5 Kaur and Abas,11 Psycharis

(2005), Haney,10 Aydin and Tasci4 the model proved to be able

to identify the level of readiness in an educational institution.

This questionnaire consists of 47 items question with details,

demographics section five the questions, e-leaning readiness

section six the questions, part of the contents of five the ques-

tions, the technology section sixteen the questions and culture

section fifteen the questions. Response questions measured using

a Likert Scale 5 for each answer choice questionnaire. The pre-

sented with Figure 1 = tangat disagree, 2 = disagree, 3 = neutral,

4 = agree, and 5 = strongly agree/

To get the value of the level of readiness, the authors calcu-

late the mean value of each variable. Based on the mean value

then, the author uses the assessment model developed by Aydin,

CH, and Tasci,4 to determine the position of the index/e-learning

readiness level (E-LRS Assessment Model) as shown in Figure 2.

This model is used, the level of preparedness. Due to the value

used by Aydin and Tasci match the researchers used a Likert scale

in obtaining data from respondents, making it easy to interpret.

Determination of respondents using non-probability sampling or

sample saturated. This means that the samples are all members of

the population, which the respondents of this study is the whole

lecturer at the FTP UNJA some 25 people.

4. DISCUSSIONIn this section discuss the results of data processing which

includes demographic data, and e-learning readiness for each fac-

tor, and e-learning readiness UNJA FTP.

4.1. Demografic

In this study, a questionnaire distributed 25 in accordance with

the number of population is the number of lecturers in FTP

Fig. 1. Assement model by Oketch.

Fig. 2. E-LRS assessment model.4

UNJA. However, of the 25 questionnaires were distributed only

17 questionnaires were returned. The rest of the questionnaire

were not returned because the lecturer is not in place.

This means that the study response rate was 68%. Based on

data for 64.71% of the respondents were female and 35.29% of

the respondents were male, as shown in Table II.

Based on the data processing that the age distribution of

respondents in Figure 3. Lowest respondents age 26 years and the

highest age 54 years with an average age of respondents 42.18.

Table III shows the distribution of respondents according to

education level. Most respondents had education level master

as much as 70.59% and the remaining 29.41 has a doctorate

education level. The fact no one FTP lecturers have a bachelor

education.

4.2. Identification of E-learning readiness

As explained on the purpose of research, that this study iden-

tifies the level of readiness of e-learning on the FTP UNJA.

Based on data obtained through questionnaires. Score calculation

is to calculate the mean value/score of respondents of each item

question. Then, to determine the degree/level of readiness will

3140


Table II. Responden by gender.

Gender Frequency Percent (%)

Male 6 35�29Female 11 64�71

Total 17 100

be compared with the value in the model ELR. From here will

be known position/index readiness of each factor of e-learning

readiness.

4.2.1. E-Learning Readiness

After the measurement of e-learning readiness, that the faculty’s

commitment to using e-learning with a mean score of 4,188. On

average lecturers have experience as a technology-based train-

ing via video or multimedia with a mean score of 3.813. In

terms of willingness to collaborate and share information through

e-learning with a mean score of 4.125. Each lecturer aware

that the design of e-learning content that is important with a

mean score of 4,063. Awareness of the importance of factors

in e-learning technology with a mean score of 4.563. And each

lecturer is ready to switch from conventional trainin way be

based e-learning with a score based around 4,125. Overall score

for e-learning readiness 4�146 > 3�4 indicates a lecturer FTP

UNJA ready for the implementation of e-learning, but with some

improvements.

4.2.2. Content Readiness Factors

Measurement readiness of e-learning content factors include

availability, satisfaction and training. Based on the data obtained

by the fact that every lectures agree there should be a teaching

material on e-learning system with a mean score of 3,500. Each

lectures also agreed to attend the training e-learning on campus

with a mean score of 3.750. Each lecture has a basic skill that

will provide conveniences for the use of e-learning with a mean

score of 3,688. Each lecturer require more training, particularly

in the development of content with the mean score. Overall for

the readiness of the mean factor scores 3828 content > 3�4 it

shows, in terms of content factors FTP lecturer ready for the

implementation of e-learning.

4.2.3. Technological Readiness Factors

Identification technology readiness factors include access to

resources (computers and internet), skills (using computers and

the internet) and attitudes towards e-learning. Based on the find-

ings of resource identification with a total mean score of 3.829,

Fig. 3. Age distribution.

Table III. Responden by education level.

Education level Frequency Percent (%)

Bachelor 0 0Master 12 70,59Doctor 5 29,41

Total 17 100

using computer and internet skills with a total mean score of

4.125, faculty attitudes towards e-learning with a total mean

score of 3.509. Overall readiness technological factors scored

3�818 > 3�4. This means of FTP UNJA technological factors

are at the ready with a few improvements. On the other hand

though attitudes towards e-learning faculty obtain total mean

score 3.509. The show is in a position ready for the implemen-

tation of e-learning. One aspect of the attitude of just getting

the mean score of 2�875 < 3�4 (not ready). This relates to the

question “Are currently making use of eLearning?” certainly not

ready for the conditions. Currently FTP UNJA not implement-

ing e-learning in learning. Only a few professors who took the

initiative to use the applications available on the market.

4.2.4. Cultural Readiness Factors

Identification readiness cultural factors include the perception

of lecturers and management support.12 Lecture perceptions

influenced usefulness and ease of use of e-learning. From the

attitude of lecturers get total score of 3�852 > 3�4. It means cul-

turally FTP UNJA be in a position ready for the implementation

of e-learning. However there is one thing to note is on lecture

perceptions about the question “I find it easy to use e-Learning

tools” for getting a score of 3�250 < 3�4 (not ready). As for

management support with a total score of 3�500 > 3�4 indicates

FTP UNJA ready for the implementation of e-learning. However,

some things need to be considered on the question “The Orga-

nizations policies have made it possible to explore eLearning”

which scored 3�06 < 3�4 (not ready). and the question “Lack of

legal provision on Intellectual property has hindered my plans to

use eLearning?” with a score of 3�063 < 3�4. This is because as

in FTP UNJA no e-learning policies and regulations on intellec-

tual property rights.

4.2.5. FTP UNJA E-Learning Readiness

After identifying the readiness factors of e-learning, content,

technology, culture and management support in the FTP UNJA

(Table IV), obtained the fact that the current FTP UNJA are at

the ready with a mean score of 3.911.

In accordance with the e-LRS models that score is declared the

FTP UNJA are at the ready with a little improvement (Fig. 4).

Overall based on data obtained on the level of readiness of the

Table IV. FTP UNJA readiness.

E-learning readiness factors Score mean

1 Technical skill and ICT knowledge 4,1462 Content 3,8283 Technological 3,8184 Cultural 3,8525 Management support 3,500

Overall score 3,911

3141


Fig. 4. FTP UNJA e-learning readiness position.

implementation of e-learning that FTP UNJA are in a ready posi-

tion with little improvement. Factors that need special attention

in the implementation of e-learning in FTP UNJA is on faculty

perceptions of the ease of use of e-learning tools. If this is left

feeling scared/or difficult will be the limiting factor.

On the other hand the management needs to formulate policies

relating to the implementation of e-learning and related regu-

lations of intellectual property rights. As standards, regulations,

SOPs and other things needed to support successful implementa-

tion of e-learning. Although lectures already have a basic knowl-

edge and use of ICT equipment, still needed some training such

as content management, audio-visual, audio visual, the develop-

ment of e-learning, pedagogy and assessment.

5. CONCLUSIONBased on the results of this study, the fact that of the factor

basic knowledge and use of ICT (computers and internet) (score

4.146), content (score 3.828), technology (score 3.818), culture

(score 3.852) and management support (a score of 3,500). On

average each factor has a value score of >3.4, it indicates FTP

UNJA be in a position ready for the implementation of e-learning

with some improvement.

From cultural factors, especially the lectures perception

towards e-learning, FTP UNJA are at the ready position with a

score of 3,852. This means that each lecture has a positive per-

ception of the use of e-learning. Which needs to be addressed

is the perception that e-learning was difficult to use, particularly

the use of e-learning tools (a score 3.250).

Although based management support has a mean score of

3,500, which means being in the ready position, but one of the

aspects that are part of management support is the support of

e-learning policies and regulations on intellectual property rights

still be in a position not yet ready.

Acknowledgments: This research is supported and funded

by STIKOM Dinamika Bangsa through an internal faculty

research grants in 2014-2015.

References and Notes1. Al-Radhi, A. Al-Din, and J. Kadhem, Bulletin of the American Society for Infor-

mation Science and Technology (Online) 34, 3 (2008).2. A. A. Al-Furaydi, English Language Teaching 6 (2013).3. A. Saekow and D. Samson, International Journal of e-Education, e-Business,

e-Management and e-Learning 1 (2011).4. C. H. Aydin and D. Tasci, Educational Technology and Society 8, 244 (2005).5. S. Borotis and A. Poulymenakou, E-learning readiness components: Key

issues to consider before adopting e-learning interventions, edited by J. Nalland R. Robson, Proceedings of World Conference on E-Learning in Cor-porate, Government, Healthcare, and Higher Education 2004, Chesapeake,VA, AACE Retrieved August, 2014 from http://www.editlib.org/p/11555 (2004).pp. 1622–1629.

6. S. Chapnick, Are you ready for e-learning, Retrieved August 2014, fromhttp://blog.uny.ac.id/nurhadi/files/2010/08/are_you_ready_e for_elearning.pdf.

7. R. C. Clark and R. E. Mayer, e-Learning and the Science of Instruction:Proven Guidelines for Customers and Designers of Multimedia Learning,Third edn., San Francisco, Pfeiffer, CA (2008).

8. T. Eslaminejad, M. Masood, and N. A. Ngah, Assessment of Instructors’Readiness for Implementing e-Learning in Continuing Medical Education inIran, University of Medical Sciences, Kerman, Iran (2010).

9. H. M. Azimi, Journal of Novel Applied Sciences 769 (2013).10. D. Haney, Performance Improvement 41, 8 (2002).11. K. Kaur, and Z. W. Abas (2004), An assessment of e-learning readiness at

Open University Malaysia, International Conference on Computers in Educa-tion (2004).

12. Oketch and H. Achieng, E-Learning Readiness Assessment Model In Kenyas’Higher Education Institutions: A Case Study Of University Of Nairobi (2013).

13. P. Sarantos, Presumptions and actions affecting an e-learning adoption by theeducational system. Implementation using virtual private networks (2005).

14. M. J. Rosenberg, The E-Learning Readiness Survey, Retrieved August2014, from http://www.books.mcgraw-hill.com/training/elearning/elearning_survey.pdf (2000).

15. P. Saunbang and P. Petocz, International Journal of E-Learning 5, 415 (2006).16. K. B. Trombley and D. Lee, Journal of Educational Media 27, 137 (2002).17. R. Vilkonis, T. Bakanoviene, and S. Turskiene, Readiness of Adults to Learn

Using E-Learning, M-Learning and T-Learning Technologies, Informatic inEducation Vilnius University (2013), Vol. 12, pp. 181–190.

Received: 30 October 2014. Accepted: 20 December 2014.

3142




Vol. 21, 3176–3180, 2015

Design Optimization Shape Web Opening of

Cellular Steel Beams

Suharjanto

Department of Civil Engineering, Janabadra University, Yogyakarta, Indonesia

The study on castellated beam with hexagon shaped web opening has been carried out, so the crack initiationat the opening-corners could easily be determined. In relation to that, the development of structural optimiza-tion with design variable is considerably needed, i.e., layout and opening size of non-angular side at I-sectionsteel beam with circular or cellular web-opening shape. Genetic Algorithm (GA) has been identified to solvedesign optimization problem. The main target is determine optimum shape, size and layout opening with cellularI-section steel beam, so the basis criterion of shape and size planning of castellated steel beam opening canbe obtained. From the data analysis of optimization using numerical computation, it can be obtained that themaximum loading capacity with even number of openings loaded in centric achieves 10%–15% compared to theodd number of openings. The castellated cellular beam of I-Section beam can increase its capacity to 3.5 timesthan the initial beam. Parametric study of the opening dimension and layout analysis with the opening heightfrom the centric with maximum cutting 0.65 hw is in the range of 0.25–0.60 from the width of openings. It canalso be known that the optimum shapes are circle and vertical ellips. It is hoped that results of this study mayfuther develop the current tradition of designing castellated steel beam.

Keywords: Castellated, Cellular Beam, Genetic Algorithm.

1. INTRODUCTIONCastellated Steel Beam (CSB) is a steel I-section beam with

opening web which is in the form of hexagon, square, circular

or modification from those three shapes. CSB used in the build-

ing and generally for pipelining installation. This will reduce the

height of ceiling from the floor and the height of the whole build-

ing. The utilization of CSB possibly replaces the conventional

way that is by hanging the pipes and ducting for air conditioning

on the beam.

Shape or geometry of the openings in the steel beams is ini-

tially hexagon, then has developed to become hexagonal shape

by adding plate in area of web-post and circular shape. Experi-

mental and theoretical studies of CSB with hexagonal openings

shape have been conducted by previous researchers to determine

the tension and deflection of the beam. Kerdal and Nethercot1

studied about the failure mode of CSB. Zaarour and Redwood2

examined the failure of the structure in CSB with hexagonal

openings shape with the adding plate in the area of web-post.

In addition, Redwood and Demirdjian5 performed research

about shear failure in web-post area of the CSB with hexago-

nal openings shape. Delesquez6 explained theorem study about

torsion at hexagon castellated beam by compiling analysis pro-

gram of elastic structure to solve the problems of torque at

castellated beams with hexagonal openings. Aglan and Redwood7

examined the failures in the area of web-post with the finite

difference method on elastoplastic behavior in the strain hard-

ening for castellated beams with hexagonal openings and it was

modified by the addition of plates on the web post.

Lawson et al.8 developed the concept strut analogy to check

the stability in the area of web-post castellated beams with circu-

lar openings. Optimization objective function is focused on the

minimum structural weight, while the design variable is repre-

sented by a geometric structure shape transverse cutting CSB.

In this paper, the Genetic Algorithm (GA) could offers good

alternative which could further improve the optimization solu-

tion. Genetic Algorithm simulating the Darwinian Theory of

Evolution.

2. LITERATURE REVIEWA 3D element to non-linear model had been developed by

Ellobody.9 This development is conducted for improving the

initial geometric imperfections, residual stresses and material

nonlinear flange, as well as web part of CSB. This study also dis-

cusses normal nonlinear analysis and high strength steel beams

cell under combined bending mode. Furthermore, Wang et al.10

elaborated the web-post buckling behavior in cellular steel beams

at high temperatures in a fire by using Finite Element Method

(FEM). A numerical model is developed to analyze the behav-

ior of web post by considering the large displacement and plas-

ticity of materials.11 Research results are very satisfactory and



can provide information merging with concrete tie-bar signifi-

cantly increase the shear resistance, ductility capacity of slip-

ping and sliding connection.10 Wang et al.11 conducted a study

on web-post buckling behavior of CSB which is fully and

partially protected at a high temperature in the fire through ver-

ified finite element models. The study states that when a web-

posting curved, the compressive tension in the web-post principle

decreases abruptly. The approach is based on European guide-

lines currently exists for the calculation of lateral-torsional buck-

ling resistance I-section beams with a calculation modification

on the cross-sectional and choice of the buckling curve.12

Based on the literature review, it can be concluded that the

study about structural optimization with variable layout design,

size of openings with the shape and cellular steel I-section beams

size at simple cellular elastic areas has not been fully developed.

This study is expected to create new novelty value on structural

optimization.

The work is quite complicated and difficult because there are

many parameters of the non-linear interaction. Thus the GA

method can be used to obtain optimal results without considering

the constraints contained in the objective function. In this study,

programming of GA optimization and tension analysis using

MATLAB programming language1�2�18 has been carried out.

3. DEFINING THE PROBLEM3.1. Problems Description

The problem faced in this research is the need for reviewing

of web post height caused by the shear forces in the web post.

Hexagonal opening of shape in the castellated beam is a fixed

geometry, whereas cellular or elliptical opening is very flexible.

Illustration of the cellular opening shape flexibility shown in

Figure 1, below.

3.2. Problems Modelling

Numerical computation simulation method had been used to

obtain the optimal shape, size, and position of I-section beam

opening with the opening web cellular. Then, the result is fur-

ther analyzed by using FEM and validated by the experimen-

tal test in laboratory. Input simulation is the profile and domain

requirements and constraints with the fitness function as follows.

Optimization problems shape, size and layout of the opening of

castellated steel cellular beam was formulated as follows:

Design variables in structural optimization of this beam is in

the form of geometry, opening dimension and distance between

opening of steel beam cellular as illustrated in Figure 1. CSB

with opening web cellular is the modification from the main

Fig. 1. Design variable of CSB.

Fig. 2. The forming process of CSB opening.

I-section beam by cutting process in circular or ellipsoid form

by overlapping as drawn at Figure 2.

From the Figure 2, the n number of openings which are formed

in the steel shown at Formula (1) and (2).

L= �n��2 ·a�+ �n+1�e = 2 ·a ·n+n ·e+e (1)

So,

n= L−e

2 ·a+e(2)

In this case, a (radius horizontal ellipses) and e (the width of

the web-post) is continuous variable or real number, while n (the

number of opening) is integer, and L (the length on CSB) is

continues.

The optimization process to define the value of n to be integer

is defined as: Value of a and e which are integers defined in

random using domain based on cutting technique by Macsteel,

a. Opening size domain, amin = 0�35 hw amax = 0�65 hw ;

b. Opening distance domain Smin = 1�08; so emin = 0�08; Smax =1�6; so emax = 0�6.

Value of ncontinues = L−e/2 ·a+e represents real number

that are rounded up to closest integer of ninteger and pro-

duces even and odd integer. Value correction of e becomes,

ekoreksi = L−2 ·a ·ninteger/ninteger +1. To obtain the maximum

area of opening, the height of openings b taken from the maxi-

mum achievement of cutting based on cutting technique is 0.65

from the height of web or hw . Therefore, the opening height of

b is constant bmax = 0�65 hw .

Fig. 3. Ellipsoid curve.

3177


0 0 1 1 1 0 1 1 0 0 1 0 1 0 1 1gen_1, 10011 gen_2, 01101 gen_3, 1101

Fig. 4. Chromosome representation.

3.3. Objective Function

The objective function is the maximum openings area with the

optimum distance between the openings of the castellated steel

beam. The area of ther opening equation can be derived as

follows:

Ellips equation,r2

a2+ z2

b2= 1 (3)

r = f �z�= a

√1− z2

b2(4)

So, ∫Ahole = L−e

2a+e

∫ z2

z1

a

√1− z2

b2dz (5)

where, b: vertical radius of the ellipse, r : abscissa of a point on

the edge of the ellipse, z: ordinate of a point on the edge of the

ellipse.

Based on previous elaboration, the objective function will be:

Maximize ZF �X�= A opening (6)

With the variable design, X = �e�a�.

3.4. Objective Function

The constraint of the inequality is in the form of the maximum

main tension that occurred ≤ allowable tension. There are two

tension occurrences. The first is the main normal maximum ten-

sion that is the highest value of the absolute value p1 and p2),

as well as the maximum friction nt�max. While the allowable ten-

sion that becomes the constraint of the friction is 0.54, for the

main tension is 0,60 y .

4. A SOLUTION ALGORITHMThis paper attempts to find the optimal solution using new

approach. Gen and Cheng15 stated that GA is the one of search-

ing algorithm that can successfully solving the model similar to

the model given in Eq. (6). GA is started from an initial popu-

lation (N). An individual in the population is known as chromo-

some which potentially represents a solution.16�17 Evaluation of

each chromosome will be measured using fitness value or objec-

tive function. The best fitness of chromosome will be selected

Parent 1

0 0 1 1 1 0 1 1 0 0 1 0 1 0 1 1

Parent 2

0 0 1 1 1 0 0 1 0 1 1 1 1 0 0 1

Off spring 1

0 0 1 1 1 0 0 1 0 1 1 0 1 0 1 1

Off spring 2

0 0 1 1 1 0 1 1 0 0 1 1 1 0 0 1

Fig. 5. Crossover operation.

Intersecting cut point

0 0 1 1 1 0 1 1 0 0 1 0 1 0 1 1

0 0 1 1 1 1 1 0 0 1

Intersecting cut point

0 1010 1

Fig. 6. Mutation operation.

from the current generation as parents to generate offspring. Each

chromosome will conduct regeneration through genetic operator

such as crossover and mutation produce offspring. Furthermore,

generation will be terminated when the optimal solution or near

optimal solution is obtained.

4.1. Chromosome Representation

According to Eq. (6), shows the structure of chromosome with

the length of 16 that is parent. In this optimization chromosome

consists of three genes that represent the variables. Genes repre-

sented in the form: string biner (10011, 01101, 11011) was show

in Figure 4.

4.2. Initial Population

The initial population will be generated randomly by a number

of chromosomes. The population is associated with the search

space to find the optimal solution. Each individual in population

may be a candidate solution to the problem.

4.3. Fitness Function

In this stage, each chromosome in the population will be eval-

uated using a measurement which known by fitness function.

Equation ZF �X� = Ahole as shown in Eq. (6) is the fitness func-

tion that to be maximize.

4.4. Selection

This genetic operator is used to select an individual chromosome

in the population with the best fitness from current generation,

and then it will become the parent for the next generation.

4.5. Crossover

Crossover operation is employed to joining each chromosome in

the population in order to produce offspring. The crossover tech-

nique includes one cut point, two cut points and multiple points

and uniform. The example of crossover operation is figured out

as follows.

Fig. 7. The convergence of GA with 200 generation.

3178


Table I. Simulation to determine weight parameter.

Load condition Number of opening Span mm Weigh parameter C1 and C2 Number of model

One point Odd 1500 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 36concentrated load and and and and and and and and and and

even 1750 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1

Two point Odd 1500 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 36concentrated load and and and and and and and and and and

even 1750 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1

Uniform Odd 1500 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 36load and and and and and and and and and and

even 1750 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1

Total model 108

Table II. Ratio of loading capacity of cellular beam.

Number of openings Ratio of loading capacity1-point loading 2-point loading Uniform loading Cellular to initial condition

Span Mm Odd Even Odd Even Odd Even Odd Even Odd Even Odd Even

1000 5 6 5 4 5 4 1.950 2.0430 1.205 1.509 1.477 1.7371250 5 6 5 6 5 6 2.348 2.4364 1.387 1.531 2.156 2.2201500 9 8 7 6 7 6 2.474 2.7992 1.536 1.940 2.298 2.7481750 9 8 9 8 7 6 2.289 2.5752 1.633 1.960 1.839 2.1762000 9 8 11 10 9 8 2.272 2.7583 1.685 2.061 2.566 2.9182250 9 10 13 12 11 12 2.191 2.3735 1.681 1.709 2.413 2.6272500 11 12 11 12 13 12 2.028 2.4339 1.640 1.758 2.417 2.845

4.6. Mutation

The purpose of mutation is to avoid local minima by prevent-

ing the population chromosomes, which may slow down or even

completely stop to process of evolution prematurely. This is

also avoid only taking the fittest chromosomes is generating the

next population, but rather adopting a weighted random selec-

tion toward that those are fitter. There is, however, one parameter

which is of vital importance in mutation, i.e., the mutation prob-

ability Pm. It control the number of mutated genes that needs

evaluation. Too small mutation probability would overlook useful

possible genes. The resulting mutated genes need to be investi-

gate for their acceptability, i.e., if they are still in the solution

domain or otherwise. A refinement process may applied to genes

which are unacceptable. The example of mutation is shown in

Figure 6.

4.7. Termination

In this step, termination of GA process may be difficult to iden-

tify convergence criteria. Based on Pasandideh et al.18�19 the cri-

teria to stop the generation is either by stopping after a fixed

number of generations or significant improvement in the solu-

tion. It is produce by comparing the average fitness value of the

current generation with that of the preceding one. This paper

conducts a fixed 200 generations to search the solution.

4.8. Population Generation

Generation of the population is resulted by the initial chromo-

some that acquires new chromosome called offspring through the

process of crossover and mutation. The descending process is

then conducted to the initial and new chromosome, followed by

determining the most fitness chromosome which is having the

most optimal value. Figure 7 show result the optimal value by

presenting the average, best and worst fitness value.

Table I, to obtain maximum fitness function takes the weight

percentage of the value of C1 (parameter area) and C2 (stress

parameters). Analysis of the simulations carried out on the open-

ing are numbered and odd that give effect to the value of C1

and C2. The objects of experiment is steel beams Profiles I elastic

area with specimens of 150×75×5×7 mm.

Table II, recapitulates the resulting carrying capacity than the

odd number of openings. More specially,

a. The increase of elastic strength is around 10%, if the even

number of openings is larger than the odd one

b. The increase in strength is around 26%, if the even number

of opening is less than the odd one.

5. CONCLUSIONS AND RECOMMENDATIONThe research reveals that

a. Optimized cellular beams with even number of web openings

subjected to load patterns considered in this research has greater

elastic carrying capacity than those odd number of web opening

b. Optimization result show that the elastic carrying capacity of

cellular beam of equal length increase sharply if the even number

of openings is smaller than that one of the odd one

c. Optimized cellular beams with even number of web openings



d. Optimized cellular beams with even number of web openings



e. Optimization result show that the elastic carrying capacity of

cellular beam of equal length increase sharply if the even number

of openings is smaller than that one of the odd one

f. Cellular opening produce by optimization have forms close to

circles for short beams and produced upright ellipses for longer

span ones

3179


g. Optimum elastic carrying capacity of cellular beams with

one point concentrated load is controlled by allowable stresses

for short span and it is controlled by deflection for long span

ones

h. Elliptical or circular cellular web opening with smooth sides

eliminate stress concentration taking place in the corners of

hexagonal openings

i. The castellated cellular beam of I-Section beam can increase

its capacity to 3.5 times than the initial beam.

References and Notes1. Anon, Partial Differential Toolbox for Use with MATLAB-User’s Guide, The

Mathwork, Inc., 3 Apple Hill Drive, Natick, MA (2006).2. A. Chipperffied, P. Fleming, H. Pohlheim, and C. Fonseca, Genetic algorithm

toolbox for use with MATLAB, User guide, Department of Automaic Controland System Engineering, University of Sheffield (2004).

3. Kerdal and D. A. Nethercot, J. Constr. Steel Res. 4, 295 (1984).4. W. Zaarour and R. Redwood, J. Struct. Eng. 122, 860 (1996).5. R. Redwood and S. Demirdjian, J. Struct. Eng. 124, 1202 (1998).6. R. Delesquez, Metallique 3, 26 (1968).7. A. A. Aglan and R. G. Redwood, Web buckling in castellated beams, ICE

Proceedings (1974), Vol. 57, pp. 307–320.

8. R. M. Lawson, Design of FABSEC Cellular Beams in Non-composite andComposite Applications for Both Normal Temperature and Fire EngineeringConditions to SCI AD 269, Fabsec Limited. (2004).

9. E. Ellobody, Thin-Walled Struct. 52, 66 (2012).10. P. Wang, X. Wang, and M. Liu, Thin-Walled Struct. 85, 441 (2014).11. S. Durif, A. Bouchaïr, and O. Vassart, Eng. Struct. 59, 587 (2014).12. S. Chen, T. Limazie, and J. Tan, J. Constr. Steel Res. 106, 329 (2015).13. P. Wang, X. Wang, M. Liu, and L. Zhang, Thin-Walled Struct. (2015).14. D. Sonck and J. Belis, J. Constr. Steel Res. 105, 119 (2015).15. M. Gen and R. Cheng, Genetic Algorithm and Engineering Optimization, John

Wily Sons, New York (2000).16. C. Saleh, V. Avianti, and A. Hasan, Optimization of fuzzy membership func-

tion using genetic algorithm to minimize the mean square error of credit sta-tus prediction, The 11th Asia Pacific Industrial Engineering and ManagementSystems Conference The 14th Asia Pacific Regional Meeting of InternationalFoundation for Production Research (2010).

17. C. Saleh, A. Hassan, B. M. Deros, M. N. A. Rahman, R. A. C. Leuveano, andA. Adiyoga, Parameters optimization of VMI system in a manufacturer andmulti retailer using genetic algorithm, International Conferance on AdvancedManufacturing and Material Engineering 2014 (ICAMME 2014) (2014).

18. S. Silva, A Genetic Programming Toolbox for MATLAB, Evolutionary and Com-plex System Group, University of Coimbra, Portugal (2004).

19. S. H. R. Pasandideh, S. T. A. Niaki, and J. A. Yeganeh, Adv. Eng. Softw.41, 306 (2010).

20. S. H. R. Pasandideh, S. T. A. Niaki, and A. R. Nia, Expert Syst. Appl. 38, 2708(2011).

Received: 27 January 2015. Accepted: 25 March 2015.

3180




Vol. 21, 3186–3189, 2015

Pathloss Calculation and Analysis Using Different

Carrier Frequency on Wideband Code Division

Multiple Access Technology

Ir. Endah Budi Purnomowati∗, Gaguk Asmungi, Anindito Yusuf Wirawan, and Rudy Yuwono

Department of Electrical Engineering, Faculty of Engineering, University of Brawijaya, Malang 65145, Indonesia

Pathloss is a decrease in the power level that caused by path attenuation of radio waves. Pathloss is dependenton several factors, one of them is the carrier frequency used. At a distance of 0.41 km and a carrier frequencyof 1920 MHz, the resulting loss is 127.41 dB. At the same distance with a carrier frequency of 2110 MHz, theresulting loss of 128.51 dB. So it can be concluded that the higher the carrier frequency used will result on thegreater the loss. To support the demand of the downlink traffic, it would be better if the carrier frequency usedin the downlink carrier frequency is lower than on the uplink.

Keywords: Pathloss, Carrier Frequency, WCDMA.

1. INTRODUCTIONPathloss is a decrease in the power level that caused by path

attenuation of radio waves, such as refraction, diffraction, reflec-

tion, and scattering. Pathloss is very dependent on the distance

of transmitter and receiver antennas and the carrier frequency

used. This study aims to provide analysis of the pathloss calcu-

lation in uplink and downlink traffic using different frequencies

according to the standard that has been designed by the Inter-

national Telecommunications Union (ITU), known as IMT-2000

(International Mobile Telecommunications 2000). It also aims to

give good recommendation on using carrier frequencies for traf-

fic downlink to support customer demands in order to keep it

running optimally. The pathloss calculation in this study using

Walfish-Ikegami model, the distance of transmitter and receiver

used are as far as 0.41 km, 0.43 km, 0.46 km, 0.67 km and

0.82 km.

2. LITERATURE OVERVIEW2.1. 3G (Third Generation)

3G, stands for third generation, is the third generation of mobile

telecommunications technology. 3G telecommunication network

services can provide support that provides rapid transfer of infor-

mation at least 200 kbps. 3G technology is the result of research

and development conducted by the ITU at the beginning of the

1980. 3G specifications and standards are developed in fifteen


years. The technical specifications are made available to the

public under the name of IMT-2000. Communications spectrum

between 400 MHz to 3 GHz are allocated to 3G.

Some application of 3G technology are:1

1. Mobile TV

2. Video Conferencing

3. Global Positioning System (GPS)

4. Location-Based Services.

2.2. Basic Concept of WCDMA

The basic concept of WCDMA is all user can communicate using

the same frequency and at the same time as well. As a result,

each user will cause interference to other users. To minimize the

effect of interference, spread spectrum method is used. To dis-

tinguish one user with the other, the random spreading code is

used. Wideband Code Division Multiple Access (WCDMA) is a

multiple access using Direct Sequence Spread Spectrum (DSSS)

technique. This technology is different from the conventional

technique that uses radio frequency bandwidth sharing techniques

available in the narrow channel into a specific slot. WCDMA

technology in data access is done continuously using certain wide

bandwidth (5–15 MHz). For each UE (User Equipment) using

services such as telephone, facsimile or multimedia, data is then

used specific codes that are correlated to each recipient. Further

discussion of WCDMA techniques will be seen from the charac-

teristics, ranging from frequency allocation, coding, scrambling,

as well as the type of modulation used.



2.2.1. Network Architecture of WCDMA

Third generation of mobile telecommunication technology or

Universal Mobile Telecommunications System (UMTS) is an

evolution of GSM, which use WCDMA radio interface. It capa-

ble of serving the transmission of data at higher speeds and dif-

ferent data rates for applications with different QoS (Quality of

Service). The overview of WCDMA network architecture can be

seen in Figure 1.

2.3. Frequency Allocation

WCDMA is designed as a standard by the ITU (International

Telecommunications Union), known as IMT-2000 (International

Mobile Telecommunications 2000). WCDMA standard set by the

so-called 3GPP (3G Partnership Project). WCDMA frequency

allocation illustrated in Figure 2.

In 3GPP, WCDMA frequencies are allocated to the arrange-

ment as Figure 2. Uplink (from the User Equipment to the

Base Station) is 1920 MHz–1980 MHz, whereas for the down-

link (from the base station to the User Equipment) is 2110

MHz–2170 MHz with a bandwidth of 5 MHz and chip rate of

3.85 Mbps.

2.4. Modulation Technique of WCDMA

The modulation technique used in WCDMA is Quadrature Phase

Shift Keying (QPSK). This modulation technique can be seen in

Figure 3.4

The signal spread and scrambled as above has a value of com-

plex chips. The real and imaginary parts are separated, where

real chips are on the branch in-phase (I) and the imaginary part

of the branch Quadrature Phase (Q).5

2.5. Pathloss

Pathloss is generally defined as a decrease in the overall

field strength corresponding increase the distance between the

transmitter and receiver. Some parameters needed to calculate

pathloss are:

1. Link Budget

2. Cell size (coverage distance of a cell).

3. Reuse distance/frequency planning.

Walfish-Ikegami method is used for pathloss calculation.

2.5.1. Walfish-Ikegami Method

L= LFS +Lrts +Lmsd (1)

For Lrts +Lmsd > 0.

Fig. 1. Network architecture of WCDMA.2

Fig. 2. Frequency allocation of WCDMA.2

Information:3

LFS is Free space loss

Lrts is rooftop to street diffraction loss.

Lmsd is multiscreen loss.

Lfs, Lrts, and Lmsd can be calculated using the following

formulas:3

LFS = 32�4+20 log d �km�+20 log f (MHz) (2)

Lrts = −16�9−10 log w �m�+10 log f �MHz�

+20 log�hm �m�+Lori (3)

Lmsd = Lbsh +Ka +Kd logd �km�

+Kf log f �MHz�−9 log b (4)

With parameters:2

MS height, hm = 1,5 m

BTS height, hb = 40 m

Building height, hr = 15 m

Building distance, b = 100 m

Road width, W = 25 m

�hm = hr −hm = 13�5 m

�hb = hb −hr = 25 m

Lori = 0�01 dB

Lbsh =−10 dB

Kd, untuk hb > hr = 7�33

Kf =−3�24.

3. RESEARCH METHODFlowchart of pathloss calculation in this study are described in

Figure 4.

Initial steps used in this study is the pathloss calculation

with a carrier frequency of 2110 MHz on traffic “downlink,”

then exchanged with the carrier frequency in traffic “uplink” at

1920 MHz.

After we calculate the pathloss at different frequencies, we

will know the great of loss is obtained. After which we can draw

conclusions on the frequency that have small losses, and can be

applied on downlink traffic.

Fig. 3. QPSK modulation in WCDMA.4

3187


Fig. 4. Flowchart of pathloss calculation.

4. CALCULATION AND RESULT4.1. Pathloss Calculation

(Uplink 1920 MHz–Downlink 2110 MHz)

By using Eqs. (6)–(8) and the parameters that have been

described previously, the calculation of the traffic pathloss

“downlink” and “uplink” at a distance of 0.41 km is

4.1.1. Distance 0.41 km

Uplink

Lfs = 32�4+20 log 0�41+20 log 1920 = 90�32 dB

Lrts = −16�9−10 log 25+10 log 1920+20 log 13�5+0�01

= 24�57 dB

Lmsd = −10+54+7�33 log 0�41−3�24 log 1920−9 log 100

= 12�52 dB

L = 90�32+24�57+12�52 = 127�41 dB

Downlink

Lfs = 32�4+20 log 0�41+20 log 2110 = 91�14 dB

Lrts = −16�9−10 log 25+10 log 2110+20 log 13�5+0�01

= 24�98 dB

Lmsd = −10+54+7�33 log 0�41−3�24 log 2110−9 log 100

= 12�39 dB

L = 91�14+24�98+12�39 = 128�51 dB

Table I. Great pathloss (uplink 1920 MHz–downlink 2110 MHz).

Distance (km) Uplink (dB) Downlink (dB)

0.41 127.41 128.510.43 127.97 129.010.46 128.78 129.880.67 133.24 134.330.82 135.64 136.74

Table II. Great pathloss (uplink 2110 MHz–downlink 1920 MHz).

Distance (km) Uplink (dB) Downlink (dB)

0.41 128.51 127.410.43 129.01 127.970.46 129.88 128.780.67 134.33 133.240.82 136.74 135.64

After getting the pathloss, using the same formulas, do the

calculation at 0.43 km, 0.46 km, 0.67 km, and 0.82 km. so we

get great pathloss depicted in Table I.

4.2. Pathloss Calculation

(Uplink 2110 MHz–Downlink 1920 MHz)

By using Eqs. (6)–(8) and the parameters that have been

described previously, the calculation of the traffic pathloss

“downlink” and “uplink” at a distance of 0.41 km is

4.2.1. Distance 0.41 Km

Uplink

Lfs = 32�4+20 log 0�41+20 log 2110 = 91�14 dB

Lrts = −16�9−10 log 25+10 log 2110+20 log 13�5+0�01

= 24�98 dB

Lmsd = −10+54+7�33 log 0�41−3�24 log 2110−9 log 100

= 12�39 dB

L = 91�14+24�98+12�39 = 128�51 dB

Downlink

Lfs = 32�4+20 log 0�41+20 log 1920 = 90�32 dB

Lrts = −16�9−10 log 25+10 log 1920+20 log 13�5+0�01

= 24�57 dB

Lmsd = −10+54+7�33 log 0�41−3�24 log 1920−9 log 100

= 12�52 dB

L = 90�32+24�57+12�52 = 127�41 dB

Fig. 5. Pathloss calculation.

3188


After getting the pathloss, using the same formulas, do the

calculation at 0.43 km, 0.46 km, 0.67 km, and 0.82 km. So we

get great pathloss depicted in Table II.

From the above calculation, graphical result are obtained as

follows:

From Figure 5, can be seen by using frequency 1920 MHz at

traffic downlink is able to produce a less loss, so that the services

provided on the traffic downlink will be better.

5. CONCLUSION AND RECOMMENDATION5.1. Conclusion

From the calculations, it can be concluded that the high carrier

frequency used in the uplink and downlink traffic, it will result on

the greater the loss. To make the downlink traffic better, then on

the downlink carrier frequency is better to use lower frequency

than the uplink carrier frequency.

5.2. Recommendation

To meet the needs of traffic on the downlink is more dense than

the traffic on the uplink, from the results of the calculations

have been done, found that the downlink traffic will have less

loss when the frequency used is lower than the frequency of the

uplink.

References and Notes1. H. Holma and A. Toskala, WCDMA for UMTS, John Wiley & Sons, England

(2001).2. K. Simanjuntak, Analysis Calculation Link Budget Indoor Penetration (WCDMA)

Wideband Code Division Multiple Acces and (HSDPA) High Speed DownlinkPacket Acces (Case Study PT. XL AXIATA Tbk.) (2011).

3. U. K. Usman, Mobile Communication Systems CDMA 2000-1x, Informatika(2010).

4. R. L. Freeman, Fundamental of Telecommunications, John Wiley & Sons,England (1999).

5. R. Vaughan and J. B. Andersen, Channels, Propagation and Antennas forMobile Communications, IEE (2003), p. 753.


3189




Vol. 21, 3190–3194, 2015

Internet Banking Transaction Authentication

Using Mobile One-Time Password and QR Code

Puchong Subpratatsavee1�∗ and Pramote Kuacharoen2

1Department of Computer Science, Faculty of Science at Siracha, Kasetsart University Siracha Campus,Chonburi, 20230, Thailand

2Department of Computer Science, School of Applied Statistics, National Institute of Development Administration,Bangkok, 10240, Thailand

Internet banking has become the norm for many simple banking transactions such as money transfers, goodsand services payments, etc. However, conducting banking transactions via the Internet may be subjected tomany types of attacks including password attacks, malware, phishing, and other unauthorized activities. Manybanks have enhanced their security by using One-Time Password (OTP) as another authentication method inaddition to traditional username and password. The OTP may be sent to the mobile phone number of the accountowner via SMS. Even with the enhanced security measure, the Internet banking is still vulnerable to differenttypes of attacks such as online phishing. We propose, design, and implement two transaction authenticationschemes using mobile OTP and QR Code. Both schemes are resilient to known attacks including, but not limitedto, eavesdropping, replay, message modification, and phishing.

Keywords: OTP, QR Code, Internet Banking, Authentication, Mobile Phone.

1. INTRODUCTIONInternet banking has become the norm for many simple bank-

ing transactions such as money transfers, goods and services

payments, etc. The user simply logs into the account using the

username and password as the credentials through a secure con-

nection. When the user requests a financial transaction such as

money transfer, and goods and services payment, the bank may

require the user to confirm the transaction using another form

of authentication. For example, the bank sends an SMS One-

Time Password (OTP)1 to the user’s previously registered mobile

phone. The user commits the transaction by submitting the OTP

through the web page. The transaction is completed when the

bank receives the valid OTP. However, making banking transac-

tions via the Internet may be subjected to many types of attacks

including password attacks, malware, phishing, and other unau-

thorized activities. The GSM network has several security vulner-

abilities. Only the airway traffic between the mobile station and

the base transceiver station is optionally encrypted with a weak

and broken stream cipher.2�3 The attacker can listen to telephone

conversions and secretly read SMS messages to commit online

crimes.

The SMS OTP is also vulnerable to online phishing attacks.

The attacker may coerce the victim to log into the phishing

site masquerading as the victim’s bank site. The phishing site


captures the username and the password of the victim and

prompts the victim to submit the OTP. Meanwhile, the attacker

uses the victim’s credentials to log into the victim’s bank account

and makes a financial transaction. The bank generates an OTP

and sends it to the victim’s mobile phone. Unaware of the fraud-

ulent activity, the victim enters the OTP and submits it to the

attacker. Subsequently, the attack submits the received OTP to

confirm the transaction to the bank. Since the OTP is identical

to the one sent from the bank, the OTP validation is valid and

the transaction is authenticated. The objective of this paper is to

present transaction authentication schemes for Internet banking

using the mobile OTP and the QR Code. The schemes must be

resilient to known attacks such as eavesdropping, replay, mes-

sage modification, and phishing. The bank sends the transaction

authentication request that is stored in the QR Code and dis-

played on the user’s web browser. The user scans the QR Code,

and the software on the mobile phone calculates the OTP for the

user to submit to the bank through the web browser to confirm

the transaction.

2. BACKGROUND AND RELATED WORKSUser authentication is the process of verifying an identity claimed

by the user.4–6 The user must provide the authentication infor-

mation to prove oneself to a verifying entity. The authentication

information may exist as, or be derived from a knowledge factor

(something the user knows), a possession factor (something the



user has), and an inherence factor (something the user is). Two-

factor authentication is an approach, which requires the user to

present two different forms of authentication information.7 The

approach decreases the probability that the user is presenting

false evidence of its identity. Lee et al.8 present an authentication

method using the mobile OTP with the QR Code. The proposed

system relies on the certificate authority to verify the OTP. How-

ever, the proposed authentication system does not protect the user

from online phishing attacks. Comparing the random number that

appears on the computer screen with the one obtained from scan-

ning the QR Code does not add any security since the attacker

can modify both values. The OTP is calculated from the trans-

action information, the perceived time, and the hash of the serial

number of the user’s mobile device. These values are not secret.

Therefore, the generation of the OTP is insecure. Moreover, there

is no explanation how the mobile device reads the transaction

information and the perceived time since these values are not in

the QR Code.

Liao and Lee9 proposed a user authentication scheme based on

the QR Code. In the verification phase, the user sends the ID and

a time stamp to the service provider. The service provider derives

the long-term secret key from hashing the user ID and its long-

term secret key. The service provider sends the QR Code contain-

ing the resulting value from performing a bitwise exclusive-OR

between the long-term secret key and a random number. The

hash of the random number, the user time stamp, and the service

time stamp are also transmitted to the user. For this scheme, an

attacker can pretend to be a valid user and can request for verifi-

cation. Thus, the attacker can obtain many QR Codes. Although

the random number is used only once, the attacker may perform a

bitwise exclusive-OR of the values in the QR Codes to obtain the

exclusive-OR of two random numbers and may be able to learn

a part of the long-term secret key. This is because the long-term

secret key is used many times. It is similar to using one-time

pad or an initialization vector in the cipher block-chaining mode

more than once. Moreover, since many values are not included in

the QR Code, the user has to type them into the mobile phone,

which is inconvenient.

We have also proposed a transaction authentication using

HMAC based OTP10 which can generate the OTP more effi-

ciently. However, the transaction information is sent in clear text.

3. PARALLEL PROVING ALGORITHM

BASED ON SEMI-EXTENSION RULEThe proposed schemes enhance the security of the current Inter-

net banking environment where the user uses username and pass-

word to log into the account. The connection between the client

and the server is done through HTTPS. After the user logs into

the system, a finance transaction can be done by entering the

transaction into the web browser and submitting to the bank’s

server. The bank’s server replies with a QR Code, which contains

relevant information regarding to the transaction, and requests the

user to enter an OTP. The user can use a mobile phone to obtain

the transaction information. The software on the mobile phone

processes the information stored in the QR Code and presents

it to the user. After visually verifying the information, the user

may confirm the transaction by entering the OTP into the web

browser and submitting it. When the bank’s server receives the

OTP, it verifies whether or not the OTP is valid. The transaction

is committed if the OTP is valid and a confirmation is sent to

the user. Otherwise, an error message is displayed on the web

browser. The designs of Internet banking transaction authentica-

tion schemes consist of symmetric cryptography and asymmetric

cryptography is described in the following subsections.

3.1. Symmetric Cryptography Scheme

For this technique, the user must register for the mobile OTP

service in order to use this method as described in Figure 1.

Registering can be done using the ATM. The bank generates a

secret key and sends the QR Code containing the secret key to

the ATM. The user may scan the QR Code to obtain the secret

key. Alternatively, the bank can send print out of the QR Code

via postal mail. This secret key is shared between the bank and

the user. The secret key is saved on the user mobile phone. Using

password-based encryption can protect this secret key. The user

supplies a password, which is used to derive a key to encrypt the

shared secret key. Before using the shared secret key, the user

must enter the correct password. The advantages for this scheme

are the transaction information is encrypted and the encryption

can be performed more quickly than the asymmetric ciphers.

The transaction confirmation process begins when the user

types in the transaction information on the web browser and sub-

mits the transaction information to the bank. After receiving the

transaction request, the bank randomly generates a nonce N1 and

computes the hash value of the transaction information TI and

N1 using a cryptographic hash algorithm. The hash value is used

to verify the integrity of the message. The transaction informa-

tion, which consists of TI, N1, and the hash value, is encrypted

using the shared secret key KA. The encrypted value is encoded

as the QR Code and sent to display on the user’s web browser.

The bank records the OTP issuing time. When the QR Code is

displayed, the user scans the QR Code and decrypts the message

with the shared secret key KA using a mobile phone. The user

computes the hash value of the received TI and N1, and compares

the hash value to the calculated value to verify the integrity of

the received message. If the verification is valid, the user inspects

the transaction information to make sure that it is accurate. The

OTP is derived from N1. The user may type the OTP in the web

browser to proceed with the transaction as illustrated in Figure 2.

Upon receiving the OTP, the bank verifies the OTP’s expiration

time. If the OTP is expired, the transaction is canceled and the

error confirmation is transmitted. Otherwise, the bank proceeds

with the verification. The bank also derives the OTP from N1

using the same technique. Therefore, the bank is able to check

the validity of the received OTP. The valid OTP implies that

the person who has knowledge of the secret key created it. That

User Mobile Phone BankATM

Register for Mobile OTP

Submit Request

Secret Key

QR Code

Secret Key

Select Scan Option

Scan QR Barcode

Fig. 1. Mobile OTP registration for symmetric cryptography scheme.

3191


User A Bank

• generate nonce N1

• compute h(TI || N1)

QR[ E(KA, TI || N1 || h( TI || N1)) ]

TI

• KA=secret key

• TI = Transaction Information

• read and decode QR Code• decrypt the message• verify the integrity of the resulting message• compute OTP = f(N1)

OTP

• check if the OTP has not been expired• compute f(N1)• verify if f(N1) is equal to OTP

comfirmation

• KA=secret key

Fig. 2. Transaction confirmation using secret key.

person was able to decrypt the transaction confirmation request

to learn the value of the nonce N1. Encrypting it with the shared

secret key KA protects the nonce N1. Only the user and the bank

know this value. Therefore, the parties with the knowledge of the

secret key can only compute the valid OTP.

3.2. Asymmetric Cryptography Scheme

In order to use this scheme, the mobile OTP registration is

required. However, the registration process for this scheme is

different from the previous two schemes. The user must have a

public-private key pair and the bank’s public key stored on the

mobile phone. The bank must have its own public-private key

pair and the user’s public key. A Certificate Authority (CA) can

be used but it is not required for this scheme to work. This tech-

nique encrypts the transaction information and does not require

shared keys. When the user registers for the mobile OTP ser-

vice, the bank presents the QR Code of its public key certificate

on the ATM. The bank’s public key can be obtained by scan-

ning the presented QR Code using a mobile phone. In Figure 3,

the user public-private key pair is generated at the ATM and is

displayed as a QR Code. The key pair can be transferred to the

mobile phone by scanning the QR Code. The user’s public key

certificate is submitted to the bank via the ATM. Besides using

the aforementioned method, the user may obtain the public key

certificate by other means. For example, the user may use a CA.

However, the key pair must be installed on the mobile phone and

the public key certificate must be submitted to the bank.

When the user desires to make a financial transaction, the user

enters the transaction information and submits it to the bank via

the web browser. The bank randomly generates a nonce N1 and

creates a digital signature on the transaction information and the

nonce using its own private key. For efficiency, a session key

Ks is generated and is used to encrypt signed information which

consists of TI, N1, and the signature. The user’s public key is

used to encrypt the session key. The bank creates the QR Code

User Mobile Phone BankATM

Register for Mobile OTP

Submit Request

Public Key CertificateQR Code

QR Code

Submit User Public Key Certificate

Bank's Public Key

Select Scan Option

Scan QR Barcode

Generate User Key Pair

Click Next

Confirmation

Select Scan Option

Scan QR Barcode

User Key Pair

Register User's Public Key

Fig. 3. Mobile OTP registration for asymmetric cryptography scheme.

of both encrypted entities and sends the QR code to the user’s

web browser. The user obtains the information stored in the QR

Code using the mobile phone. The session key can be obtained

by decrypting the encrypted session key with the user’s private

key. Then, the encrypted information can be decrypted using the

session key. Subsequently, the user verifies the bank’s signature.

If it is valid, the OTP is derived from the nonce N1. Similar to

the two previous schemes, the user must inspect the transaction

information before submitting the OTP to confirm the transac-

tion. The server checks if the OTP is received within the allowing

period. The unexpired OTP is verified by performing the same

computation on N1. The transaction confirmation process for this

scheme is shown in Figure 4. For this technique, the user can

be certain that the received message is from the bank by verify-

ing the digital signature. The bank can be assured that the valid

OTP is from the user who is performing the transaction request

because the encrypted message can only be decrypted with the

private key corresponding to the public key. The user learns the

nonce and uses it to generate the OTP.

4. SECURITY ANALYSESThe proposed schemes use two-factor authentication in which the

password is used to enter the transaction and the key is used to

create an OTP to commit the transaction. We assume that the key

distribution step is done securely. The communication between

the bank and the user’s web browser is secure using HTTPS

protocol. The attacker cannot obtain the information from eaves-

dropping. The eavesdropper cannot break the HTTPS connec-

tion within the limited period. However, the information flow

between the user’s mobile phone and the web browser is inse-

cure. It assumes that the user securely stores the long-term keys,

namely; secret key and private key, on the mobile phone. The

following subsections discuss possible attack types.

4.1. Eavesdropping

Eavesdropping is an unauthorized real-time listening to private

communication. The objective is to obtain information that is

being transmitted. An attacker may be able to capture the QR

3192


User A Bank

• generate nonce N1• create signature on TI || N1• generate session key Ks• X1= E(Ks, TI || N1|| signature)• X2= E(ae, Ks)

QR[ X1, X2]

TI

• ae, ad = user A’s key • be = bank’s public key• TI = Transaction Information

• be, bd = bank’s key • ae = user A’s public key

• read and decode QR Code• decrypt X2 using ad• decrypt X1 using Ks• verify signature• compute OTP = f(N1)

OTP

• check if the OTP has notbeen expired

• compute f(N1)• verify if f (N1) is equal to OTP

comfirmation

Fig. 4. Transaction confirmation using public key.

Code from the user’s computer screen. However, the attacker will

not be able to obtain the secret key or the private key informa-

tion since there is no information regarding the key. The attacker

should not be able to decrypt the encrypted information without

the secret key or the private key. Therefore, the valid OTP should

not be generated.

4.2. Message Modification

A message modification attack is an assault on the integrity

of a security system, in which the attacker intercepts mes-

sages, then alters or reorders them to produce an unauthorized

effect. There are two scenarios where the attacker could modify

the transaction information for personal gain. The first scenario

involves modifying the transaction information sent to the bank

and the second scenario involves modifying the QR Code. For

the first scenario, the attacker intercepts the transaction request

The transaction confirmation request(a) (b) The transaction verification result

Fig. 5. Screen captures of the implementation.

and modifies it. For instance, the attacker changes the receiv-

ing account to another account. After transaction information has

been modified, it is submitted to the bank. Correspondingly, the

bank generates the QR code and sends it to the user request-

ing the OTP confirmation. Automatic validation by the software

would be deemed valid. However, when the user inspects the

transaction information, it would be different from the actual

request.

For the second scenario, if the attacker could intercept the

communication and could modify the QR Code, the information

contained in the QR Code would be different from the one sub-

mitted by the user. However, the integrity check would fail. Each

message contains an integrity check using cryptographic hash

code, or digital signature depending on the technique.

4.3. Replay

A replay attack occurs when an attacker repeats a valid data

transmission or delays the original transmission. The aim is to

compromise the integrity of the system. The attacker may be

able to masquerade as the actual party who are making the data

transmission or may cause adverse effects on the system.

All schemes employ the use of the randomly generated number

or the nonce. Therefore, the OTP is unique to the transaction.

Furthermore, the OTP can be used only once within a time limit.

When the user requests a financial transaction, the bank saves

the requested transaction. A nonce and the transaction request

time are also recorded. The transaction is open awaiting the OTP.

If the OTP is received, the transaction is completed. A replayed

confirmation that arrives after the transaction is closed does not

have an effect. On the other hand, a replayed OTP that arrives

during the transaction period does not match the one calculated

by the bank. Hence, the replay attack is infeasible.

4.4. Phishing and Man-in-the-Middle

Phishing is a type of Internet fraud that attempts to acquire a

user’s credentials by deception. The attacker sets up a website

that masquerades as a trustworthy site. Phishing messages are

sent to lure to the user to visit the site. Unknowingly, the user

enters the username and password, which are captured by the

phisher.

Consider the phishing attack scenario where the attacker

entices the user to log into the bank’s site through the attacker’s

site. The attacker may be able to obtain the user’s password.

3193


However, the attacker should not be able to make a transaction

since it requires the valid OTP, which was derived from the nonce

using the key, or protected by the key. Suppose that the user

also performs a transaction through the attacker’s website. For

example, the user transfers some money to another account, and

the attacker modifies the transaction information to transfer the

money to the attacker’s account and sends the transaction request

to the bank. The bank sends the QR Code to the attacker and the

attacker presents the QR Code to the user on the computer screen.

When the user scans the QR Code using the mobile phone, the

verification will pass. However, the user will notice that the trans-

action information is different from the one previously requested.

The user will decline to enter the OTP. Without the OTP confir-

mation, the transaction is incomplete. Online phishing is still a

threat since it requires the user to inspect the transaction infor-

mation before entering the OTP. However, a vigilant user will

not be a victim of such an attack.

5. EXPERIMENTAL RESULTSWe implemented the prototypes of the proposed schemes. The

application on the mobile phone is an Android application and

the web application is written in JSP. ZXing library is used for

generating and reading the QR Code. For security API, we use

Java Cryptographic Architecture (JCA). HMAC-SHA256 is used

for generating the message authentication code. AES and RSA

algorithms are used for symmetric and asymmetric cryptographic

operations, respectively. Figure 5 shows the screen capture of

the transaction confirmation request after the user submits the

transaction information. The QR Code contains the transaction

confirmation request, which the user needs to use the key to

create a valid OTP. The OTP for the transaction is valid for three

minutes. Otherwise, the transaction will be canceled. When the

user captures the QR Code using the mobile phone with the

Mobile OTP software, the transaction information is verified and

the OTP is calculated.

The transaction information, the OTP, and the verification

result are displayed as illustrated in Figure 8. The user can visu-

ally verify the transaction information and may submit the OTP

to the bank to commit the transaction.

6. CONCLUSIONSThe transaction authentication using mobile OTP and QR Code

can be implemented in a secure manner. We presented three

authentication schemes namely; symmetric cryptography, and

asymmetric cryptography, that can be securely used to authenti-

cate the user’s transaction. Both schemes can prevent attacks such

as eavesdropping, message modification, replay, and phishing.

For preventing online phishing attacks, the user must visually

inspect the transaction information to ensure its integrity before

submitting the OTP.

References and Notes1. A One-Time Password System, IETF RFC 2289-1998.2. E. Barkan and E. Biham, Conditional estimators: An effective attack on A5/1,

SAC 2005, LNCS, edited by B. Preneel and S. Tavares, Springer, Heidelberg(2006), Vol. 3897, pp. 1–19

3. A. Biryukov, A. Shamir, and D. Wagner, Real time cryptanalysis of A5/1 ona PC, FSE 2000, LNCS, edited by B. Schneier, Springer, Heidelberg, (2001),Vol. 1978, pp. 1–18.

4. Securing mobile devices: Present and future, http://www.mcafee.com/us/resources/reports/rp-securing-mobile-devices.pdf.

5. Internet Security Glossary, IETF RFC 2828-2000.6. W. Stallings, Cryptography and Network Security: Principles and Practice,

6th edn., Prentice Hall, Upper Saddle River, NJ (2013).7. R. J. Boyle and R. R. Panko, Corporate Computer Security, 3rd edn., Prentice

Hall (2012).8. Y. S. Lee, N. H. Kim, H. Lim, H. Jo, and H. J. Lee, Online bank-

ing authentication system using mobile-OTP with QR-code, Proc. 5th Int.Conf. Comput. Sciences and Convergence Information Technology (2010),pp. 644–648.

9. K. Liao and W. Lee, Journal of Networks 5, 937 (2010).10. P. Subpratatsavee and P. Kuacharoen, Transaction authentication using

HMAC-based one-time password and QR code, Computer Science and itsApplications, LNEE (2015).


3194




Vol. 21, 3215–3219, 2015

Hybrid Supply Chain Operations Reference-System

Dynamics Performance Measurement MTS-MTO

Production Typology for Batik Industry

Taufiq Immawan1�3�∗, Marimin2�∗, Yandra Arkeman2, and Agus Maulana3

1Department of Industrial Engineering, Faculty of Industrial Technology, Universitas Islam Indonesia2Department of Agroindustrial Technology, Faculty of Agricultural Technology, Bogor Agricultural University, Bogor, 16002, Indonesia

3School of Business and Management, Bogor Agricultural University, Bogor, Indonesia

Batik is an Indonesian cultural asset that should be preserved. In the perspective of its protection, supply chainperformance is important to be assessed, especially in its industrial aspect. Supply Chain Operations Reference(SCOR) is a method of self-assessment and a comparison of activities in the supply chain performance evalua-tion using five attributes; reliability, responsiveness, costs, assets management, and agility. Measurements withSCOR method presents a framework of business processes, performance indicators, best practices, as wellas a unique technology to support communication and collaboration inter-supply chain, so as to improve theeffectiveness of supply chain management and the effectiveness of supply chain improvement. SCOR focusedon causality of the steps that are shaped linear correlation. This state of condition appears to be the weaknessof SCOR method, because it cannot continue more specific steps forward that can predict performance. In thisstudy, SCOR in combination with System Dynamics (SD) can identify the four perspectives and their interactionvariables associated with causal loop design model. The design of a hybrid model SCOR-SD provides moreeffective and interesting value. Furthermore, the simulation of the model from data inputs in 2014 creates predic-tion until 2019. The results obtained for the reliability attributes associated with perfect order fulfillment startedin 2015 to 2019 respectively 85.16% to 91.06%. Responsiveness attributes associated with the order fulfillmentcycle time started at 69.71 days to 62.04 days. The simulation results shown that the associated total cost wasfluctuated ranging from IDR 1,491,041.700 to IDR 1,470,531,891. Attributes associated with cash managementassets to cash cycle time in a row is 35 days to 26 days. The latter attribute is an attribute associated with theagility of supply chain flexibility upside respectively start from 2.26 % to 1.8 days.

Keywords: Supply Chain Operations Reference, Batik Industry, System Dynamics, Performance.

1. INTRODUCTIONMeasurement of the company’s supply chain needs to be done to

keep the distribution of products continuously. Since its introduc-

tion in 1996, the majority of researchers using the supply chain

operations reference (SCOR) to measure the performance of the

supply company.4

There are 5 attributes Introduced in SCOR as a measure of

supply chain performance assessment, namely reliability, respon-

siveness, cost, asset management, and agility. All attributes could

be broken down into several levels which then becomes a metric.3

Supply chain performance measurement needs to be conducted

to determine the performance level of the supply chain, so that

it can be further improved. SCOR model is a tool used to map

and improve supply chain operations.4 In the research also men-

tioned that the SCOR model has been highlighted in Sweden

over the last few years and several companies are now adopting

∗Authors to whom correspondence should be addressed.

the SCOR model in their supply chain operations. Since its intro-

duction in 1996, SCOR experienced some major revisions based

on practical needs.2 SCC3 introduced the SCOR version 11.0, in

which there are five attributes that can be used as a benchmark

to assess the performance assessment in the supply chain, the

attributes of reliability, responsiveness attributes, cost attributes,

attribute management assets, as well as agility attributes that can

later be in-breakdown into levels found in SCOR metrics appro-

priate type of production. Previous research8 elaborated a litera-

ture review to result a solution by using the SCOR model, then

to use Analytical Hierarchy Process (AHP) analysis for target

process selection. AHP can aid in deciding which supply chain

processes are better candidates to re-design in light of predefined

criteria. The results provided by the AHP analysis go beyond

target process selection. From the AHP analysis, it can calculate

a priority rank for the metric criteria used.

Research on supply chain performance improvement using the

SCOR has been carried out by experts or by the industry itself.



In the USA has conducted a computer-aided supply chain con-

figurations based on the SCOR model. In this study contribute

to the tool-assisted configuration and development2�10 also con-

ducts supply chain performance measures that contribute to the

performance of Chinese furniture manufacturing supply chain.

System Dynamics (SD) is one method that can solve com-

plex problems with simulation models. It is known that the prob-

lems are considered complex systems due to the components

that are inside interact dynamically and provide causation. Such

interaction is called the Causal Loop. Interactions that occur

can be modeled in the form of a mathematical model, which in

turn can be calculated with the help of a computer in the form

of simulation.1 Research using primary approach was made by

Ref. [8], 20075 have conducted an analysis of the value chain by

using a hybrid simulation and AHP. Simulations were carried out,

using an amalgamation of discrete event simulation with con-

tinuous simulation (System Dynamics). However, less obvious

SCOR contribution with SD (System Dynamics) in this study. It

is based on the perspective of the SCOR model is not based on

five attributes in SCOR. According to Ref. [3] this research pre-

sented the development of a dynamic of sustainable supply chain

for Gayo coffee for improving the sustain ability performance.

The dynamic model is proven from the comprehensive descrip-

tion and analysis of the SGCSS systems operation taking

into economic, social and environmental dimension. Economic

dimension focused on actors profits, social dimension empha-

sized on the actors skills and environmental dimension concerned

on the pulp utilization as compost.

The outline of this paper to measure supply chain performance

using a combination of two methods of SCOR and SD, herein

after referred to as Hybrid SCOR and SD to improve dynamic

performance in the batik industry with the MTS-MTO produc-

tion type in Indonesia. Contribution this research are provide the

model for corporate using combination between SCOR perfor-

mance that static measurement and System Dynamic. This com-

bination can be simulated, to predict the final result.

2. LITERATURE REVIEW2.1. Supply Chain Operations Reference (SCOR)

Approach

2.2. SCOR Definition

According to Ref. [4] model of Supply Chain Operations Refer-

ence (SCOR) was delivered by the Supply Chain Council (SCC),

which was established in 1996. The SCOR model created by the

SCC aimed to provide a method of self-assessment and compar-

ison of the activities and performance supply chain as a standard

cross-industry supply chain management. This model presents a

framework of business processes, performance indicators, best

practices (best practices) as well as a unique technology to sup-

port communication and collaboration intra-supply chain, so as

to improve the effectiveness of supply chain management and the

effectiveness of supply chain improvement. Meanwhile, accord-

ing to Ref. [2], SCOR model is a process reference model, which

is intended to make the industry standard that enables supply

chain management in the next generation. Assessing the perfor-

mance parameters such as asset management, cost, reliability,

agility and responsiveness, does evaluation. SCOR performance

section consists of two types of elements: Performance Attributes

and Metrics.4

2.2.1. SCOR Five Performance Attributes

According to Ref. [4] in the supply chain performance measure-

ment through the method of SCOR version 11.0, there are five

attributes of work that will be done to measure the performance,

while the five attributes are:2.2.2.1. Reliability. Reliability is an attribute that is focused

on the consumer. A supply chain should be consumer centric and

companies in the supply chain are need to meet consumer needs.

Reliability stated ability to perform tasks expected. Metric of

reliability consist of right quantity, right in time, right in quality.

Performance indicator is perfect order fulfillment.2.2.2.2. Responsiveness. Responsiveness attribute states how

quickly a task is executed. This shows consistent speed in run-

ning the business. Primary performance indicator is order fulfill-

ment cycle time. Responsiveness focuses on the consumer.2.2.2.3. Agility. Agility attribute states ability to respond to

external changes like the ability to change. External influences

are unexpected demand increase or decrease, stopping operates

from suppliers, disaster, terrorism act, or employee problems.

Primary performance indicator are flexibility and adaptability.

Agility focuses on the consumer.2.2.2.4. Cost. Cost is an attribute that internal focus.

Attributes charge states the cost of running the process. Costs

generally include the cost of labor, raw materials, transportation

costs. Primary performance indicator is total cost to serve. Total

cost focuses on the consumer because of measuring the total cost

to serve the consumer.

2.2.2.5. Asset Management. Asset management attribute states

the ability to efficiently utilize assets. Asset management strategy

in the supply chain include inventory reduction and determine

whether in source or out source productions. Primary perfor-

mance indicator are cash to cash cycle time and return on fix

asset. Asset management focused on internal.

The reason for using 5 attributes are: reliability is used to mea-

sure the level of the fulfillment of order, whether can be right

100 percent or still underneath it. As for the responsiveness is

used to measure the velocity of process level from order accepted

until the order is shipped. The level of responsiveness compare

with the target of the companies. Agility used to measure the flex-

ibility of production, as a reaction over the rise or decline in the

order with 20 percent range. Cost is used to calculate the total

cost to serve so that can be known by the efficiency of the process

of batik production. The value of the total cost to serve compared

with the target company. Asset management used to measure cash

to cash cycle time. Duration time from corporate make payment

for the order, delivering product, until customer cash release.

2.2.2. SCOR Metrics

SCOR metric model includes 134 metric level 1. By using a hier-

archical approach as developed in the process of SCOR, metric

also has several different levels.

(1) Metrics diagnostic level 1 is the overall health of the supply

chain. This metric is also known as strategic metrics and key

performance indicators (KPI). First level of benchmarking met-

ric helps companies set realistic targets to support the strategic

direction.

(2) Metric Level 2 acts as a diagnostic for the metric level 1.

Relationship diagnostic help identify root causes of performance

gaps in the metric level 1.

(3) Metrics level 3 acts as a diagnostic for 2-level metrics.

3216


Fig. 1. Causal loop model from five SCOR performance attributes.

2.3. System Dynamics (SD) Approach

SD is a methodology for studying and managing feedback of

variables contained in the complex system1 said the primary

method of studying the problem with a systematic viewpoint,

where the elements of the system interact with each other in

a relationship given feedback to produce a behavior. The inter-

action in this structure is translated into a mathematical model

with the aid of a digital computer simulated to obtain historical

behavior. SD models can create a feedback to decision makers

about the possible absence of the collision of a series of wisdom

to simulate and analyze the behavior of the system on different

assumptions.

Fig. 2. Flow diagram model from five SCOR performance attributes.

2.3.1. Causal/Feedback Loop

Causal loop diagrams or also known as influence diagrams,

are used to help modelers to understand the system by provid-

ing an overview through causality in the system (the system

conceptualization).

Causal loop diagram give simplicity in understanding the

structure and behavior of a system, however the simplicity of it

can also make unclear whether the relationship between variables

happened is the relationship or not rate to level.5 By using the

Causal Loop Diagram modelers can quickly structuring model

based on the assumptions used as in Figure 1. Causal loop dia-

gram used to help model maker understand the system by giving

a general picture via relationship of cause and effect in the

3217


Table I. The final result of simulation.

No Attribute Actual Initial simulation Final simulation Target

1 Reliability 85.58% 85.16% 91.06% 100%2 Responsiveness 91.02 days 69.71 days 62.04 days 80 days3 Agility 2.5 days 2.26 days 1.8 days 1.8 days4 Cost IDR 1,482,637,610 IDR 1,491,041,700 IDR 1,470,531,891 IDR 1.400.000.0005 Asset management 32.73 days 35 days 26 days 30 days

system. By causing loop model maker can quickly draw up a

structure model based on assumptions used.5

2.3.2. Flow Diagram

Flow diagram is a representation of a form of detailed depiction

of the system, as in Figure 2. In the flow diagram shown vari-

able types and kinds of relationships between variables in the

system. The main objective of the flow diagram is to represent

the flow and structure of the system in detail in order to facilitate

mathematical modeling.

A diagram describing the relationship between variables made

in the loop diagram case and effect with clear, where used

symbols in certain variables. In the flow diagram distinguished

between the flow of physical and the flow of information. Change

in a variable in this sub system will change physical quantity.

Otherwise the flow of information is not the flow of a convertible.

The information derived from one source could be variable

that transformed to another without reducing the amount of infor-

mation that is in the source.

3. RESEARCH METHODCausal loop that has been created in Figure 1, converted into

flow diagrams and processed using Powersim 9 software. Flow

diagram represent the mathematical model together with the data

processing variables was constructed and described in Figure 2.

4. RESULTS AND DISCUSSIONAttributes associated with perfect reliability orders fulfillment

simulation run for 5 years from 2015 to 2019, following the sim-

ulation results obtained respectively 85.16%, 85.19%, 87.97%,

89.02%, and 91.06%. From the simulation results can be known

company’s performance can be quite adequate to see the feed-

back that occurs within or between attributes in SCOR. However,

it should be done to improve performance in the future until

100%.

Responsiveness attributes associated with the order fulfillment

cycle time, respectively 69.71 days, 67.37 days, 65.49 days,

64.75 days and 62.04 days. From the simulation, the results

are through the target, are 80 days, better with state of reality,

but the problem of how to minimize the cycle time of service

may be considered for the future. 1,485,517,574, IDR 1, 479,

374,070, IDR Attributes associated with the total cost consecu-

tive associated with the total cost consecutive service charge IDR

1,491,041,700, IDR 1,472, 708, 030 346, and IDR 1.470.531.891.

Attributes associated with cash management assets to cash cycle

time in a row is 35 days, 33.2 days, 31.62 days, 28.66 days, 26

days. The later attribute is an attribute associated with the agility

of supply chain flexibility upside respectively 2.26 days, 2.03

days, 1.98 days, 1.88 days and 1.8 days.

Its better if we use the weighted of each parameter like Analyt-

ical Hierarchy Process (AHP). The systems approach was accom-

plished by identifying all of the factors contained in the system

to obtain a good solution for resolving the problem, and then

creating a model of AHP to help rational decisions. AHP model

is used to calculate the weight of criteria, both quantitative and

qualitative in one research.

Graphically, AHP decision problem can be constructed as

a multilevel diagram (hierarchy). AHP begins with the focus

or goal past the first level criteria, sub criteria, and finally

alternative.6 The final result of simulation can be seen in Table I.

There are some better result between before and after the

simulation. There are some significant improvement because of

lean manufacturing and customer order decoupling point between

Make To Stock and Make To Order.

The managerial implication from this result are the SCOR

parameter hybrid with dynamic model make the model of system

dynamic are able to equip the result of SCOR having the charac-

ter of static. Hybrid SCOR-SD it helps manager of the company

understand interactions among SCOR, of the parameters of so

that by improving one or two of variable will affect the whole of

variable. Manager just looking for variables that have potential

biggest then fix it. The end result is the fifth variable will also

change.

5. CONCLUSION AND SUGGESTION5.1. Conclusion

Reliability for the most influential attribute variable is on sched-

ule production orders (100%), the smallest influence is fulfillment

of orders to customers (58.77%) compared to among the four

main supporting variables.

The most influential attribute for responsiveness is production

cycle time (48.59 days), the smallest influence variable is delivery

time (1.03 days).

The most influential variable in Cost attribute is in the pro-

duction cost and the smallest influence is from procurement vari-

ables. The highest number for asset management is number of

days outstanding loans (57.52 days) and the smallest is number

of variables influence the pending sales (1.03 days).

Meanwhile, the most influential Agility variables is supply

source flexibility (83.33%) and the smallest influence is the make

variable supply flexibility (12.83%).

5.2. Suggestions

It is suggested to apply Analytical Hierarchy Process (AHP) for

weighting the SCOR attributes for measuring the overall supply

chain performance.

Acknowledgment: The Authors would like acknowledged

to School of Business and Management, Bogor Agriculture

3218


University, Indonesia and Directorate of Academic Development,

Universitas Islam Indonesia, Indonesia.

References and Notes1. J. W. Forrester, Industrial Dynamics, Massachusetts Institute of Technology,

Massachusetts-U.S.A: The M.I.T. (1961).2. S. H. Huang, S. K. Sheoran, and H. Keskar, Computer and Industrial

Engineering 48, 377 (2005).3. J. R. Machfud and R. S. Marimin, Journal of Theoretical and Applied Infor-

mation Technology 70 (2014).4. J. Paul, Transformasi Rantai Suplai Dengan Model SCOR (Cetakan ke-1

edn.), edited by R. Nurul, A. D. Zalsa, H. Wahyudi, A. Rosyid, and T. Erlinda,Jakarta Pusat, Penerbit PPM (2014).

5. K. E. Maani and R. Y. Cavana, System Thinking and Modelling UnderstandingChange and Complexity, Prentice Hall, New Zealand (2000).

6. D. M. A. Marimin, P. M. P. I. F. P. Machfud, and B. Wiguna, Journal of CleanerProduction 85, 201 (2014).

7. J. A. Palma-Mendoza, International Journal of Information Management 34,634 (2012).

8. F. Persson, International Journal of Production Economics 131, 288(2011).

9. L. Rabelo, H. Eskandari, T. Shaalan, and M. Helal, International Journal ofProduction Economics 105, 536 (2007).

10. D. J. Robb, X. Bin, and T. Arthanari, International Journal of Production Eco-nomics 112, 683 (2008).

11. V. Salton and S. H. Belton, Adding Value to Performance Measurementby Using System Dynamics and Multicriteria, Strathcyde Business School,Research Paper No. 2001/19 (2001).

Received: 23 October 2014. Accepted: 15 December 2014.

3219




Vol. 21, 3240–3243, 2015

Continuous Review Probabilistic Inventory

Analysis Using Type-2 Fuzzy Logic

Muhammad Ridwan Andi Purnomo

Department of Industrial Engineering, Faculty of Industrial Technology,Universitas Islam Indonesia, Yogyakarta, Indonesia

A practical and novel approach for probabilistic inventory analysis is presented in this paper. In uncertaintyenvironment, where demand of every period is not constant, then the risk to have stock out can be occurredduring ordering lead time. In this study, since the inventory can be monitored every time, hence, continuousreview technique is proposed as the basis of analysis. The nature of continuous probabilistic inventory analysisis very complicated. Even though it can minimise inventory cost, however, such technique is not practical forindustrial world due to its complication. To model probabilistic factors, besides mathematical model, Fuzzy Logic(FL) can be used for that purpose. In FL, probabilistic factors that cause uncertainties will be represented usingFuzzy sets. However, in conventional FL, the Fuzzy sets are developed using precise curves and sometime itcannot cope with uncertainties. Therefore, in this study, a Type-2 FL is proposed to model probabilistic factorsin the inventory analysis. A case study shows that the proposed model able to give reasonable solution to theproblem.

Keywords: Continuous Review Model, Inventory, Probabilistic, Type-2 Fuzzy Logic.

1. INTRODUCTIONIn industrial world, demand fulfilment is the main factor to

generate profit. However, since there are uncertainties in the

operational level, such as machine breakdown, un-experienced

workers, lateness of raw material arrival and so on, it will make

demand cannot be fulfilled well. Hence, inventory of raw materi-

als or finished goods are required to anticipate such uncertainties.

On the other hand, inventory will cause extra investment and it

will reduce product competitiveness due to its high price. There-

fore, inventory must be controlled well in order to avoid stock

out and high investment.

There are several established techniques to control the inven-

tory. In general, there are two models, which are deterministic

and probabilistic inventory model.

Deterministic inventory model is used to control inventory

under constant demand environment while probabilistic model

is used to control inventory under dynamic demand. Focus of

this study is on probabilistic model since the demand is vary-

ing from period to period. In term of probabilistic model, there

are two techniques that usually used to control the inventory,

which are periodic and continuous review. In periodic review

model, the inventory will be observed at equal interval of time,

hence, the decision variables in such technique is the interval

of time and the maximum inventory. When the inventory can

be observed continuously, continuous review model will be the

suitable technique. Decision variables in such technique is order

quantity and reorder point.

Theoretically, determination of order quantity that can min-

imise total inventory cost, further it called as Economic Order

Quantity (EOQ), is carried out through mathematical mod-

elling and usually the model is very complicated. Even though

the mathematical model can provide optimum result, however,

it will be impractical for industrial world due to its complica-

tion. In industrial application, Research and Development (RND)

staffs, will decide the order quantity based on their experiences.

By considering several factors, they will decide the order size

intuitively and it is more practical. However, if their experiences

can be systemised, then the decision will be more consistent and

precise. Human opinion usually is mixed with ambiguity and

uncertainty and such factors cannot be incorporated in formal

method such as mathematical model. Therefore, a technique that

able to accommodate ambiguity and uncertainty is required.

In engineering field, there is a technique that deal with ambigu-

ity and uncertainty, it called as Fuzzy Logic (FL). Such technique

has been applied widely in several fields such as engineering,

operational research, decision analysis and management. In FL,

ambiguity and uncertainty is modelled using Fuzzy sets and

Fuzzy rules. Fuzzy sets usually are developed using Fuzzy curves

and, in conventional FL, the curves are precise, there is no

tolerance or interval for every Fuzzy curve. Sometime, precise

Fuzzy curves cannot accommodate experts opinion. Researchers

in Fuzzy field have proposed Type-2 FL (T2FL) to overcome

disadvantages of conventional FL. In T2FL, every Fuzzy curve

has an interval. It is believed that the interval can accommodate



experts opinion and it makes T2FL more acceptable by the

experts.

This study elaborates application of T2FL to model probabilis-

tic inventory system. The case is continuous review with back

order. Uncertainty factors to be considered are average demand,

demand during lead time and inventory cost components.

2. THEORETICAL BACKGROUND2.1. Inventory

The main objectives of inventory analysis is to find the answers

of how much should be ordered and when should the order be

placed so that the total inventory cost can be minimised. Inven-

tory cost can be classified as the cost of carrying inventories

(holding cost), the cost of stock out (opportunity cost) and the

cost for replenishing inventories (order cost). Total carrying cost

is cost to carry a unit inventory multiplied by number of inven-

tory per period. Total stock out cost is cost to handle stock out,

that can be back order or lost sales, multiplied by number of

stock out unit while total order cost is cost to place an order

multiplied by order frequency to fulfil demand.

The investigated industry in this study is applying back order

to postponed demand. Inventory cycle in back order environment

is shown in Figure 1. Initially, the inventory level is full (Imax).

The inventory level will be decreasing from time to time because

of demand fulfilment. When the inventory level reaches a cer-

tain point, then an order will be placed. Such point is called

as reorder point (r�. While waiting the order to come, demand

during that period (lead time/l) will be fulfilled using remain-

ing inventory stock. Stock out (E�x�� might be occurred when

there is no inventory stock anymore to fulfil demand. Inventory

level will be updated by order quantity (Q� when the order was

received.

Hence, total cost of inventory is formulated as follows1:

TC�Q�r� = AD

Q+h

(Q

2+ r −Dl

)+ �D

QE�x� (1)

Reorder point can be determined based on following equation.∫ Dl

r

1

Dldx = hQ

�D(2)

while following equation is used to determined E�x�.

E�x�=∫ Dl

r�x− r�f �x�dx (3)

where: Q: order quantity, r : reorder point, A: ordering cost per

order, D: average annual demand, h: holding cost per unit per

period, Dl: expected demand during lead time, �: back order

cost per unit per period, E�x�: expected back order quantity per

period.

Fig. 1. Inventory cycle.1

2.2. T2FL

In conventional or Type-1 FL (T1FL), membership value of a

crisp input in a Fuzzy set is crisp value. Sometime, such tech-

nique can not anticipate the effect of ambiguity and uncertainty.

Recently, T2FL which is use interval value for membership value

of a crisp input, has been widely applied to overcome the limita-

tion of T1FL. Figure 2 shows the difference between Fuzzy set

in T1FL and T2FL.

From Figure 2, in T1FL, if crisp input 25 is mapped to the

‘Low’ Fuzzy set, then the membership value will be 1. It means

25 is always ‘Low.’ However, other experts may have different

perceptions. All experts may be agree that 25 is ‘Low’ but the

membership value will fall in interval 0.67 to 1. T2FL can be

used to accommodate such different perception.

As shown in Figure 2 above, Fuzzy set of T2FL has an inter-

val. Therefore, a crisp input will have two membership value

namely upper membership value and lower membership value.

In T2FL, Fuzzy rules are created based on IF-THEN rule and it

is same with T1FL. When Mamdani style is applied as the infer-

ence system, then there will be two output domain which are

lower output domain and upper output domain. Such condition

is depicted in Figure 3.

In order to get centre of gravity of the output domains, an ap-

proach to define shifting points (the red circle) from upper to

lower and from lower to upper membership value is required. In

this study, Karnik-Mendel algorithm2 was used for that purpose.

2.3. Related Works

In term of inventory management, continuous review technique

has been developed. In a study, application of continuous review

technique in a supply chain has been investigated.3 In that study,

a warehouse with m-identical retailers is analysed by consider-

ing average annual order frequency, expected number of back

order and budget constraints. Since the problem become non-

linear integer programming, hence, a parameter-tuned Genetic

Algorithm (GA) is proposed as the solution searching algorithm.

The proposed model is successfully used to manage inventory in

the investigated supply chain system.

Basic model of continuous review is for single item. The devel-

opment of such technique for multi item has been investigated by

previous researcher.4 In that study, besides for multi product, the

continuous review technique is also being used to analysis multi

period production planning. Since the variable is growing expo-

nentially, then a GA is used as the optimisation technique. The

study shows the proposed GA has relatively better performance

compared to Integer Linear Programming developed by Lingo

software.

Another study has investigate mixture back order and lost

sale case in continuous review technique.5 Fuzzy annual aver-

age demand and Fuzzy demand during lead time is considered.

Since the environment is mixed with uncertainty, then the authors

proposed an intelligent algorithm based on Fuzzy simulation is

Fig. 2. Difference between T1FL and T2FL.

3241


Fig. 3. Lower and upper output domain.

proposed as the model solving algorithm. The proposed model is

proven capable in solving the faced problem.

Several success applications on T2FL in engineering will be

elaborated in the following sections. T2FL which combined with

modular Artificial Neural Network (ANN) face recognition has

been investigated.6 There are 2 T2FL applied, the first T2FL

was used for feature extraction in the training data while the

second T2FL was used to predict the relevance of the modular

ANN as the recognition module. The study shows that FLT2 can

increase the performance of modular ANN by facilitating the

representation of human expert perceptions.

T2FL also has been applied in modelling of multi echelons

supply chain system.7 The FLT was used to model several param-

eters in the supply chain such as forecast demand, inventory level,

transportation distances, transportation cost, stock out level, stock

out cost, carry over and holding cost. For decision variables opti-

misation, a Genetic Algorithm (GA) has been used. Such study

shows that T2FL has better performance in representing uncer-

tainties compared to T1FL.

In term of prediction, T2FL also has been applied for carbon

monoxide concentration forecasting.8 In that study, footprint of

uncertainties of Fuzzy sets are extracted by implementation of an

interval type-2 Fuzzy C-Means (FCM) algorithm and based on an

upper and lower value for the level of fuzziness m in FCM. The

study shows that T2FL has superior performance in comparison

with T1FL.

Based on related works above, it can be concluded that T2FL

is similar with T1FL. However, if there is vagueness in the input

variables of a system, then T2FL be more suitable to model the

system compared to T1FL.

3. MODEL DEVELOPMENTIn continuous review probabilistic inventory model, the input

are average annual demand, demand during lead time, holding

and ordering cost. The output is order quantity (Q) and reorder

point (r). In this study, in order to get optimum r , Q value must

be determined first. Therefore, the first step in the modelling is

providing Fuzzy set for every input and Q variable as the out-

put variable. In this study, normalised value for every variable

is applied. The real interval of average annual demand, demand

during lead time, holding, ordering cost and Q is 50–150 units,

0–25 units, 50–180 per unit per month, 100.000–350.000 per

order, 50–150 units, respectively. Figure 4 shows the Fuzzy set

for every variable after normalisation.

After Fuzzy sets development, the next step is Fuzzy rules

determination. Since the proposed T2FL is an expert system,

then Fuzzy rules are defined based on expert opinion. There are

15 Fuzzy rules used to represent the condition of the inventory

environment. All of the rule use ‘AND’ as the conjunction in the

antecedent part since all of the input variables must be considered

Fig. 4. Fuzzy set for every input variable.

together in order to determine the output. Table I shows the Fuzzy

rules.

Since every Fuzzy set has 2 curves namely upper and lowercurve, then every input value will be mapped to upper and lower

curve. Hence, every Fuzzy rule will have firing value in the form

of an interval. Equation below shows formula to calculate lower

and upper limit of the firing range.

�lfrn�ufrn�=( N∏

n=1

�xn↓ ×�xn↓N∏

n=1

�xn↑ ×�xn↑

)(4)

where: lfr = lower firing rule, ufr = upper firing rule, n = rule

index, �x=membership value of input x, ↓= lower Fuzzy curve,

↑= upper Fuzzy curve.

To evaluate how the proposed model works, an input vector is

required. Let say, i = �0�60�0�58�0�48�0�63� is the input vector,

then the result of Fuzzy rules evaluation is as shown in Table II.

By applying Karnik-Mendel algorithms, shifting points are

obtained, with upper to lower (ul) point is 8 and lower to upper

(lu) is 8. Hence, output from ul and lu can be calculated as

follows:

yul =�∑8

i=1 fi ↓×yi ↑�+ �∑15

i=9 fi ↑×yi ↑�∑8i=1 fi ↓+

∑15i=9 fi ↑

(5)

yul = 1

ylu =�∑8

i=1 fi ↑×yi ↓�+ �∑15

i=9 fi ↓×yi ↓�∑8i=1 fi ↑+

∑15i=9 fi ↓

(6)

ylu = 0�4

Table I. The fuzzy rules.

Rule D Dl h A Q

R1 H H H H M (0.4–0.6)R2 H M N H H (0.7–1.0)R3 H M N N M (0.4–0.6)R4 M H H H M (0.4–0.6)R5 L M H N L (0.2–0.4)R6 L L H N L (0.2–0.4)R7 L M N N M (0.4–0.6)R8 M M N N M (0.4–0.6)R9 H L N H L (0.2–0.4)R10 H M N H H (0.7–1.0)R11 M H H N M (0.4–0.6)R12 L L N N L (0.2–0.4)R13 H L H N L (0.2–0.4)R14 M L H H M (0.4–0.6)R15 L M N H M (0.4–0.6)

Note: H = high, M =medium, L= low, N = normal.

3242


Table II. Firing of every fuzzy rule.

Rule Firing (f ) Average firing �f � Consequent

R1 [0.02695, 0] 0.013475 [0.4, 0.6]R2 [0.231, 0] 0.1155 [0.7, 1.0]R3 [0, 0] 0 [0.4, 0.6]R4 [0.0294, 0] 0.0147 [0.4, 0.6]R5 [0, 0] 0 [0.2, 0.4]R6 [0, 0] 0 [0.2, 0.4]R7 [0, 0] 0 [0.4, 0.6]R8 [0, 0] 0 [0.4, 0.6]R9 [0, 0] 0 [0.2, 0.4]R10 [0.231, 0] 0.1155 [0.7, 1.0]R11 [0, 0] 0 [0.4, 0.6]R12 [0, 0] 0 [0.2, 0.4]R13 [0, 0] 0 [0.7, 1.0]R14 [0, 0] 0 [0.4, 0.6]R15 [0, 0] 0 [0.4, 0.6]

Finally, Q as output can be calculated as follow:

Q = y = yul+ylu2

(7)

Q = 0�7

The Q value above is still normalised value. When it is con-

verted to the real Q value, then 120 units is obtained. The r value

can be calculated as follows. Assuming � = 500.∫ Dl

r

1

Dldx = hQ

�D(8)

∫ 25

r

1

25dx = 115×120

500×100

r

25= 1−0�276

r = 18

And E�x� = −5�9 units. Since E�x� is less than 0 then back

order will not occurred. Total cost for the decision proposed by

T2FL is as follows.

TC�120�18� = 225�000×100

120+115

(120

2+18−25

)+0

TC�120�18� = 193�593

4. DISCUSSIONFuzzy logic usually is developed by experts. The proposed T2FL

is also developed by the experts who have wide experiences about

the investigated inventory system. Therefore, the result can be

considered as feasible solution for the investigated system. How-

ever, T2FL is not an optimisation tool, the result provided by the

proposed T2FL may be not the best solution. Total cost of the

solution may be not the most minimum one.

The advantage of the proposed T2FL is the ability to cope with

uncertainty of input variables. In optimisation model of inventory

system, which is based on mathematical model, such uncertainty

is not responded. In real industrial applications, uncertainty will

frequently occurred and T2FL will be more applicable compared

to the mathematical model.

5. CONCLUSION AND SUGGESTIONIn T2FL, fuzzy set for every input variable has an interval. In

the consequent part, output also modelled in the form of interval.

Such condition is believed can cope with uncertainties. Based on

the explanation above, it is proven that the proposed T2FL can be

used to model continuous review probabilistic inventory system

with backorder case.

For the next study, it is proposed to hybrid the proposed T2FL

with an optimisation algorithm such as Genetic Algorithms (GA).

The optimisation tool can be used to adjust parameter values

inside the T2FL so that an optimum solution can be obtained.

References and Notes1. E. A. Elsayed and T. O. Boucher, Analysis and Control of Production System,

Prentice-Hall, Inc., New Jersey (1994).2. J. M. Mendel, Uncertain Rule-Based Fuzzy Logic Systems: Introduction and

New Directions, Prentice-Hall, Upper Saddle River, NJ (2001).3. S. H. R. Pasandideh, S. T. A. Niaki, and N. Tokhmehchi, Expert Systems with

Applications 38, 11708 (2011).4. I. Saracoglu, S. Topaloglu, and T. Keskinturk, Expert Systems with Applications

41, 8189 (2014).5. L. Wang, Q.-L. Fu, and Y.-R. Zeng, Expert Systems with Applications 39, 4181

(2012).6. O. Mendoza, P. Melín, and O. Castillo, Applied Soft Computing 9, 1377 (2009).7. S. Miller and R. John, Knowledge-Based Systems 23, 363 (2010).8. M. H. F. Zarandi, M. R. Faraji, and M. Karbasian, Applied Soft Computing

12, 291 (2012).


3243




Vol. 21, 3249–3253, 2015

Physicians’ Acceptance of Electronic Health

Records Exchange: An Extension of the with

UTAUT2 Model Institutional Trust

Malik Bader. Alazzam∗, Abd. Samad Hasan Basari, Abdul Samad Sibghatullah, Mohamed Doheir,Noorayisahbe Mohd Yaacob, and Farah Aris

Faculty of Information and Communication Technology, Universiti Teknikal Malaysia Melaka (UTeM),76100 Durian Tunggal, Melaka, Malaysia

Electronic health records (EHRs) exchange improves hospital quality and reduces medical costs. However, fewstudies address the antecedent factors of physicians’ intentions to use EHR exchange. Based on institutionaltrust integrated with UTAUT2 model, we propose a theoretical model to explain the intention of physicians touse an EHR exchange.

Keywords: EHR, UTAUT1, UTAUT2, Institutional Trust Physician, Acceptance.

1. INTRODUCTIONFor nearly two decades, the medical industry began to evolve

through the exchange of health information system where they

were provided with comprehensive coverage of medical care in

developed countries, the majority of patients tend to visit numer-

ous hospitals throughout their survives, and “hospital shopping”

has become a relatively common occurrence. There-fore, the

majority of patients tend to visit several hospitals throughout their

lives, and “hospital shopping” has converted a relatively com-

mon comparatively. In fact, hospitals in developed country could

have avoided considerable medical resource wastage if they had

exchanged patient information the issuance of these countries

(IC) card is issued by health sector for patients. It includes an

embedded miniature IC chip with various medical certificates.

Residents can not only conveniently receive medical services but

also inquire health insurance situation and other medical-related

records. The Healthcare Certification Authority (HCA) issues the

Physician IC card for each certificated physician a physician

IC card. Only physicians can access and retrieve the patient’s

medical information with the physician IC card. Therefore, the

Department of Health (DOH) was tasked with establishing an

electronic health record (EHR) exchange platform, where by hos-

pital and medical staff used certification-authority IC cards with

public key encryption. Physicians can search for patient anam-

neses by using the BNHI IC card, the physician IC card, and

electronic medical certificates on the Internet. EMR exchange

has played a crucial and central role in health care by providing


patient information that supports numerous health care applica-

tions, such as the diagnosis, treatment, and prevention of disease.

For these information-technologies (IT)-enabled benefits to man-

ifest in developed countries, physicians must first adopt EHR

exchange systems. EHR exchange adoption is an instance of

information system (IS) acceptance and use in a setting that com-

bines IS adoption with health care elements and thus requires

distinct theorization within IS literature. Despite an emerging

benefit in the field of health informatics, few factors in hospi-

tals’ adoption of EHR exchange have been identified,2–5 and only

a limited and fragmented understanding of physician behavior

exists on EHR contextual exchange.1 Physician EHR-exchange

behaviour has certain variations compared to usual user behavior.

The EHR exchange provides a lifesaving mechanism for health

care, which is a type of service.

EHR exchange must occur among competing hospitals. There-

fore, EHR exchange is not a normal activity, but a socioeco-

nomic interactive operation between health care organizations

and the environment in which they exercise. The concerns of

physicians about the adequate functioning of an IS application

(patient privacy, security) are likely to stop the diffusion of infor-

mation, such as in EHR exchange. To build physician trust, EHR

exchange providers must do more than merely providing elec-

tronic linkages.

These differences identify risks and the uncertainty of an

online environment. The significant importance of trust should

be emphasized in the EHR exchange. Existing variables of the

technology acceptance models do not fully reflect the motives

of users; therefore, it is necessary to identify additional intrinsic

motivational factors. Previous research indicated that the need for



incorporating additional factors improved the predictive ability

and explanatory power of the behavioral intention.8�9 A litera-

ture review reveals that further research is necessary to clarify

the role of trust and risk perceptions in physicians’ acceptance

of eHealth.

Based on UTAUT2 theory and healthcare literature, we have

developed an extended version to predict the behavior of physi-

cians. We achieve this by framework to know which fac-

tors affected on physician behavior; we add institutional trust,

perceived risk, and their most critical antecedents. Our study

aims are:

(a) This framework to examine whether institutional trust and

perceived risk significantly affect behavioral intention (BI) of

physicians’ use of an EHR exchange system;1

(b) To investigate whether the effect of institutional trust on BI

is not only direct, but is also achievable, by reducing perceived

risk;

(c) To clarify which factor has significantly influences on deci-

sions to use EHR exchange; and

(d) To evaluate whether the extension of UTAUT2 can pro-

vide a better explanatory power to predict the adoption of EHR

exchange.2

1.1. Literature Review

Reference [1] a survey by Shapiro et al. indicated that 97% of

physicians believed that health information exchanges improve

healthcare safety and quality. The application of ISs among

physicians has been a critical research topic in the field of health

informatics. Previous studies have investigated the factors influ-

encing IS acceptance by physicians’ use of several information

technologies, such as bedside computer technology,14 and EHR.3

Compared with previous studies, we specifically highlight the

factors driving physicians’ intention to use an EHR exchange

system.

1.2. EHR Exchange

EHR exchange systems electronically transferred patient level

clinical, demographic and health-related information between

disparate hospitals. These exchanges offer various opportuni-

ties to achieve the following six improvements to patient care:

(a) Safety; (b) Effectiveness; (c) Patient centeredness; (d) Time-

liness; (e) Efficiency; (f) And equity.7 Without a mechanism to

exchange EMR among hospitals, the inability of hospital staff

to review the medical history of patients who have visited other

hospitals could also result in redundant.4

However, the implementation of an EMR exchange system

requires considerable effort. Prior research has identified cer-

tain factors of EMR exchange adoption and implementation

among hospitals,5 use experiences regarding health information

exchange (HIE),6 the clinical application of HIE,7 HIE-related

privacy and security concerns], and public attitudes toward HIE.8

It is critical to encourage physicians to use the EMR exchange

system. Despite this urgent need, scant studies have explored

the antecedent factors of physicians’ intentions to use EHR

exchange. In a typical medical setting, the characteristics of the

users, technology, norms, and context may substantially differ

from those in business settings; thus, this study explored the

antecedent factors of physicians’ intentions to use EMR exchange

from the perspective of information system adoption.

1.3. UTAUT Model Review

1.3.1. UTAUT1

Reference [9] academics have showed technology acceptance

studies for over two decades now. They have used numerous con-

cepts and models to carry out these papers in different contexts

with different part of study. Findings from these researches vary.

The authors of UTAUT model unified eight concepts and mod-

els which contain Theory of reason Action (TRA) Technology

acceptance model (TAM), Motivational model (MM), Theory of

planned behavior (TPB) combined TAM and TPB (C-TAM-TPB)

Model of PC Utilization MPCU. Innovation Diffusion Theory

(IDT) and Social Cognitive.9�10

Theory (SCT) Bandura (1986). The unification by the inves-

tigators sum up all the concepts from the eight models to four

elements, which expects intentions, usage, and four moderators

of the key relationships.11 Figure 1 explains the relationships that

exist in the UTAUT model. The model has four EV, which refers

to exogenous variables, EE, which refers to effort expectancy,

PE which indicates to performance expectancy, SI which refers

to social influence, and FC which mea facilitating conditions.

The endogenous variables are the technology intention to use

and behavior. There are other four moderators age, namely, gen-

der, experience, and voluntariness. Performance expectancy is

famous, as a degree individual believes in the benefit of the sys-

tem to performance.9�12�13

The degree of ease linked with the use of the system is an

important indicator towards technology intention to use which

calls effort expectancy. The degree of an individual perceives on

the important of new system used is also significant indicator

towards technology intention to use. The degree of an individual

believes on the effective of organizational and technical infras-

tructure exists that needs to support the use of the system is an

important indicator which called facilitating condition.

The chosen of this model in this paper is justified by its univer-

sal and integrative way, incorporating a vast variety of explanatory

variables from the main theoretical models developed to illustrate

technology acceptance and use. In particular, Morris et al.11 car-

ried out an in-depth test of literature on this matter and proposed

a unified model that merge the contributions collective to the pre-

vious theories. Therefore, it is reasonable to expect a theory that

integrates the most important contributions from other models to

be more to the previous theories explanation of technology accep-

tance and use.

1.3.2. UTAUT2

Reference [14] extends the unified theory of acceptance and use

of technology (UTAUT) to investigate acceptance and use of

technology in a consumer context. That the goals of UTAUT2

integrates three concepts into UTAUT: HM, PV, and HT the

demographic characteristics of service users’ were used as mod-

eratos variables namely experience, age and gender to control the

effect on the BI and the use of technology. The findings have

derived from two-channels online survey conducted with user of

technology. The data collected took four months from 1,512 of

the clients of mobile. As compared to UTAUT, the additions pro-

posed in UTAUT2 produced a substantial improvement in the

variance explained in BI.

3250


Fig. 1. UTAUT2 conceptual model.14

1.4. Hedonic Motivation (HM)

Hedonic motivation (HM) can defined as the intrinsic motivation

such as fun, enjoyment or pleasure when using a technology

because of technology for its own sake, and it has known an

important construct in determining technology acceptance and

use. HM is similar to perceived enjoyment or playfulness to TAM

as an intrinsic motivation factor.14–16

1.5. Price Value (PV)

In general, people chose the services or products when their

benefit gives more than the price value compared with its cost.

Therefore, price value can be defined as learners’ cognitive trade-

off among the seeming benefits of the applications and the eco-

nomic cost.14�16�17

1.6. Habit (HT)

Habit (HT) is one of a strong predictor of future technology

use.17 Habit has been identified as the degree to which individuals

incline to implement behaviors routinely due of learning.9�14�15

1.7. Institutional Trust

Physicians have limited ability to monitor or control the EEC

use of their EHR exchange, which is why trust is required.

Institution-based trust exists when trust is associated with the

existence of third-party structures that are independent of dyadic

actions. Shapiro41 described institutional trust as the belief that a

trustor has regarding the security of a situation because of guar-

anteed safety procedures and other structures.1 Zucker indicated

that institutional trust is the most crucial trust-creating mode

among impersonal economic environments, where a sense of a

community with similar values is lacking.

Institutional trust measures were employed to address physi-

cian perceptions regarding whether the national framework

(Legal and regulatory framework, Third-party guarantees, Inter-

national standards, Directives, Escrows) was conducive to using

an EHR exchange system. Regarding technology, the EEC facil-

itates using e-signatures to replace physical signatures or seals

on medical records in accordance with the law for verifying the

identity of a signee. EHR security can be enhanced using certified

encryptions to protect the privacy of patients and the integrity

of medical records, and requiring certified decryptions to access

records.1

Regarding legality, the DOH continues to amend its regulations

of EHR production and management for medical institutions; this

regulation promotes self-management and continual improvement

among EHR exchanges. Concerning policy, the DOH formulates

EHR inspection mechanisms, using the International Organiza-

tion for Standardization (ISO) 27001 guidelines to formulate

inspection items in accordance with the regulation of EHR pro-

duction and management by medical institutions. The EEC pro-

vides certified EHR logos, enabling members of the public to

identify which hospitals have adopted certified EHR and pro-

moting EHR exchanges among hospitals. Regarding standards,

the EEC provides standardized templates based on the specifica-

tions of EHR exchanges, ensuring the appropriate EHR content

to minimize infringement risks to patient privacy. Therefore, in

this study, institutional trust is a critical research

1.8. Research Methodology

1.8.1. EHRs and UTAUT2

The UTAUT model has been widely used in the EHR adoption

and acceptance which is shown in Table I. The employees will

find the EHR system (physicians) useful if it helps them to per-

form the functions of the Directorate efficiently and effectively.

PE, EE, SI, HD, PV (price value has been excluded due deferent

related this study area), HB will directly affect the intention to

use of the EHRs by the officers and staff. Thus, a high level of

intention to use is likely to increase employee adoption of EHRs.

H1. Performance expectancy is positively related to physician

intention to use EHR system.

H2. Effort expectancy is positively related to physician intention

to use EHR system.

H3. Social influence is positively related to the physician inten-

tion to use EHR system.

H4. Facilitating conditions is positively related to physician the

intention to use EHR system.

3251


Fig. 2.

H5. Hedonic motivation is positively related to physician inten-

tion to use EHR system.

H6. Habit is positively related to physician intention to use EHR

system.

H7. trust institutional is positively related to physician intention

to use EHRs.

2. FUTURE RESEARCH AND LIMITATIONSThis paper is in its first hypothetical idea, in which a beginning

model is proposed based on the previous studies and theoretical

thought. The following phase is the application and validation

of the model to a arrangement of healthcare professionals, in

order that to test the established framework and directly measure

its explanatory and predictive power. Coming studies may eval-

uate other relationships that were not expected in this model and

that will develop the ability to describe the dependent variables.

Therefore, this paper opens up other selections for future research

Refinement of the constructs and measures is one of the possi-

bilities. Additional option is the examination of more complex

relationships between the Independent and dependent variables

of the model. Testing this model with other e-health technolo-

gies, and in other countries that may be more or less developed

than developed countries in e-health use are options that can also

bring benefit.

3. CONCLUSIONSUnderstanding the acceptance and use of EHR system of physi-

cians should bring strong benefits for the future sustainability of

the Healthcare System, which will enjoy more efficient use of

resources. Therefore, the goal of this paper is to detect a set of

determinants of acceptance of EHRs by physicians. To realize

this goal, we suggest a research model based on UTAUT2, adding

new constructs trust institution. We designate this new set of con-

structs “e-health extension to UTAUT2.” We also suppose this

paper to provide a theoretical framework that is a foundation and

a starting point for future research on the acceptance of EHRs

by physicians.18

Acknowledgment: This paper is part of Doctor of philoso-

phy (Ph.D.) work in UteM and part of the work also supported

under grand PJP/2013/FTMK (17A)/S01230.

References and Notes1. P.-J. Hsieh, Int. J. Med. Inform. 84, 1 (2015).2. M.-P. Gagnon, E. K. Ghandour, P. K. Talla, D. Simonyan, G. Godin,

M. Labrecque, M. Ouimet, and M. Rousseau, J. Biomed. Inform. 48, 17(2014).

3. K. Jammoul, H. Lee, and K. Lane, Understanding Users Trust and the Mod-erating Influence of Privacy and Security Concerns for Mobile Banking: anElaboration (2014), Vol. 2014, pp. 1–11.

4. M. N. Herian, N. C. Shank, and T. L. Abdel-Monem, Health Expect. 17, 784(2014).

5. C. Kanger, Evaluating the Reliability of EHR-Generated Clinical OutcomesReports: A Case Study (2014), Vol. 2.

6. S. Trang and S. Zander, Dimensions of Trust in the Acceptance of Inter–Organizational Information Systems in Networks: Towards a Socio-TechnicalPerspective (2014).

7. J. F. Cohen, Trust, Risk Barriers and Health Beliefs in Consumer Acceptanceof Online Health Services (2014), pp. 1–19.

8. A. Balaid, M. Z. A. Rozan, and S. N. Abdullah, Asian Soc. Sci. 10, 118 (2014).9. M. D. M. B. Alazzam, Dr. A. S. Sibghatullah, and A. S. H. Basari, Eur. Sci. J.

10, 249 (2015).

3252


10. A. A. Taiwo and A. G. Downe, The Theory of User Acceptance and Use ofTechnology (UTAUT): A Meta-Analytic Review of Empirical Findings (2013),Vol. 49.

11. M. G. Morris, M. Hall, G. B. Davis, F. D. Davis, and S. M. Walton, User accep-tance of Information Technology: Toward Aunified View 1 (2003), Vol. 27,pp. 425–478.

12. M. J. Wills, Examining Healthcare Professionals Acceptance of ElectronicMedical Records Using Utaut (2008), Vol. IX, pp. 396–401.

13. A. Hennington and B. D. Janz, Information Systems and Healthcare XVI:Physician Adoption of Electronic Medical Records: Applying the UTAUT Modelin a Healthcare Context Records: Applying the UTAUT Model in A (2007),Vol. 19.

14. V. Venkatesh, Consumer Acceptance and Use of Information Technology:Extending the Unified Theory (2012), Vol. 36. pp. 157–178.

15. M. Kang, B. T. Liew, H. Lim, J. Jang, and S. Lee, Emerging Issues in SmartLearning (2015), pp. 209–216.

16. M. Technologies, S. D. Impact, and M. I. Usage, Mobile Technologies andServices Development Impact on Mobile Internet Usage in Latvia MobileTechnologies and Services Development Impact on Mobile Internet Usage inMobile Technologies and Services Development Impact on Mobile InternetUsage in Latvia (2013).

17. A. Raman and Y. Don, Int. Educ. Stud. 6 (2013).18. J. Tavares and T. Oliveira, Electronic Health Record Portal Adoption by Health

Care Consumers Proposal of a New Adoption Model (2013), pp. 387–393.

Received: 17 January 2015. Accepted: 20 February 2015.

3253




Vol. 21, 3275–3278, 2015

Performance Evaluation of Distance Measurement

in Biometric Finger Knuckle Print Recognition

Guruh Fajar Shidik∗, Syafiq Wardani Dausat, Rima Dias Ramadhani, and Fajrian Nur Adnan

Dept. Computer Science, Universitas Dian Nuswantoro, Semarang 50131, Indonesia

Biometric identification is a security method that utilize parts of human body. Finger Knuckle Print (FKP) is oneof many parts in human that relatively unique. This research try to find out the best distance measurementas recognition method with implication of pre-processing CLAHE. Several variants of distance measurementthat evaluated in this research are Euclidean, Manhattan, Minkowski, Canberra, Chi-square, Chebyshev andBray Curtis. Besides that, feature extraction Principal Component Analysis (PCA) was applied as texture featureafter preprocessing phase. The experiment results has shown several recognition that used pre-processingCLAHE able to improve accuracy with distance measurements in recognize FKP, moreover, several distanceshowed degraded accuracy. Highest accuracy that used pre-processing CLAHE are gathered when applied withChi-Square distance until 95.15%.

Keywords: Biometric, Finger Knuckle Print, Principal Component Analysis, CLAHE, Distance Measurement.

1. INTRODUCTIONBiometric authentication is a technology used to identify the

human based on physiologic and behaviouristic attributes.1 Bio-

metric refers to the unique physiologic and behaviouristic char-

acteristics which can be used to differentiate between some

individuals’ identities effectively.2�3 There are various hand print

patterns that can be used as the human identification including

finger print, hand palm print, hand form feature, hand venous

structure, and finger knuckle print.45

Between various hand based biometrics, the skin pattern

around the finger back called Finger Knuckle Print (FKP), has

the highest ability in differentiating the human’s identity.2 The

unique patterns on the finger back surface has greater potential

for special identifications based on each individual which are

then able to give proper contributions in the biometric method.6

Woodard and Flynn was the first authors introducing the fin-

ger knuckle print method as a biometric property by captur-

ing images in 3D.7 In their research, they carried out a new

approach in identifying and verifying personal identity by uti-

lizing the 3D-finger surface feature as a biometric identification.

They utilized 3D images on various types of human hands with

the images of index, middle, and little fingers which can be cal-

culated and used as a comparison in determining the similarities

from several subjects. The feature extraction

The research conducted by Zhang et al.,2 built a set of equip-

ment to capture the FKP images, and introduced FKP identifica-

tion algorithm which was highly efficient to process the real-time


data. Local convex from the FKP images was extracted based

on its local coordinate system to harmonize and cut the images

off for the feature extraction. They proposed Gabor Filtering to

conduct the feature extraction.

The research conducted by Ozkaya et al.,8 applying discrimi-

native common vectors (DCV) method to obtain an unique fea-

ture vector called Euclidean distance measurement as a proper

strategy to complete the identification and verification tasks. The

results of the identification had the 100% accuracy in the exam-

ination data.

Zhang et al.9 conducted a research by finding new sets of

equipment that could be used to capture the finger back pattern

and conduct ROI extraction based on the captured images by

determining the image axial line. The ROI images were obtained

from two FKP and then the algorithm conformation was con-

ducted to find the similarities. Then, Yu et al.,10 conducted a

research by extracting the features on the finger knuckle print

by dividing the ROI images from FKP to subblocks set from the

local features. The process of conforming the two FKP images

was conducted by calculating histogram distance intersection

between the registered and input images.

Kumar dan Ravikanth11�12 proposed other approach for per-

sonal authentications using 2D finger-back surface imagery.

They developed a system to capture the hand-back image and

then being extracted on the finger-back area using several pre-

processing steps. The analysis methods used were PCA, LDA and

ICA to conduct the characteristic extraction and conformation.

Based on the previous researches, no one tries to conduct

research of measure performance evaluation by searching for



the best results using some distance measurement in FKP. We

motivated to conduct an evaluation of the distance measurement

for FKP recognition using several distance measurements, such

as Euclidean distance, Manhattan distance, Minkowski distance,

Chebychev distance, Chi-square distance, Canberra distance and

Bray Curtis distance. This research will evaluate the use of Con-

trast Limited Adaptive Histogram Equalization (CLAHE) on the

Finger Knuckle Print pre-processing area, and this research also

examine the feature extraction using Principal Component Anal-

ysis (PCA) with 60, 100, and 120 feature reduction values to

conform the results of the extraction on the examined and exer-

cised images.

This research consists of several parts: Part two, three and four

explains about the proposed approaches. Part five describes the

experimental steps and design. Part six discusses the research

analysis and results, and the last part is conclusions.

2. CONTRAST LIMITED ADAPTIVE

HISTOGRAM EQUALIZATION (CLAHE)Contras Limited Adaptive Histogram Equalization (CLAHE) is

a technique to improve the image quality by utilizing histogram

score limit to reduce the excessive brightness and contrast of

an image.13 CLAHE is a great method in improving the signal

components and the noise from an image,14 and the obtained

formula is explained as follows:

g = g min+[

2�∝2� ln

(1

1−P�f �

)]0�5

(1)

where g min is a minimum pixel value, P�f � is a cumulative

probability distribution and � is a nonnegative real scalar spec-

ifying a distribution parameter. In this study, clip limit is set to

0.01 and � value in Rayleigh distribution function is set to 0.04.

3. PRINCIPAL COMPONENT

ANALYSIS (PCA)Principal Component Analysis (PCA) was firstly used to repre-

sent the human’s face as an image reduction and feature extrac-

tion technique developed by Sirovich and Kirby (Sirovich and

Kirby 1987) and used by Turk and Pentland15 to detect and iden-

tify the human’s face.15

4. DISTANCE MEASUREMENTThe distance measurement is a technique to conform two images.

There are several ways in determining the distance value, they

are: Euclidean distance, Manhattan distance, Mahalanobis dis-

tance, Minkowski distance, Chebyshev distance, Chi square dis-

tance, Bray Curtiz distance.

4.1. Euclidean Distance

This metric measurement is the most common way to measure

the similarities in capturing the images because of its efficiency

and effectively. It measures the distance between two image vec-

tors by calculating the root square from the square absolute dif-

ference and can be calculated:16

de��a�b��c�d�� =√�b−a�2 + �d−c�2

(2)

4.2. Manhattan Distance

Manhattan distance has a resistance towards the outlier. This dis-

tance metric is calculated by determining the number of absolute

difference between two image vectors and can be calculated.16

The Manhattan distance measurement is between the two points

(i and j) in p-dimension which is formulated as follows:17

dab =p∑

c=1

�xac −xbc� (3)

4.3. Canberra Distance

City block distance metric rises great values and make some dif-

ferences for two similar. Oleh thus, each feature with the different

pairs will be normalized by dividing them with other pairs. Can-

berra Distance is used to conduct the numeric measurement of

the distance between query vectors and database features:17

d�ab�=∑ �a−b��a�+ �b� (4)

4.4. Minkowski Distance

This general forms rise other distance metrics for positive values

of p, for instance, p= 1 it could be city block and p= 2 could be

Euclidean distance. In the metric distance comparison, we also

take the value of p= 3 as the Minkowski distance:16

d�ab�=∑��a−b�p�1/p (5)

4.5. Chebyshev Distance

Chebyshev distance or called maximum value distance is a dis-

tance comparing 2 vector values by selecting the biggest vector

value to be the value of Chebyshev distance:16

d�ab�= max��a−b�� (6)

4.6. Chi Square Distance

Chi square distance is a distance used to measure the distance

between two vectors. The basic concept of Chi Square distance

calculation is taken from Chi Square statistics used to calculate

the conformation between the distribution and frequency:18

d�ab�= 1

2

∑ �a−b�2�a+b�

(7)

4.7. Bray Curtiz Distance

It is a normalization method which is commonly used for biol-

ogy, ecology, and other environmental sciences. The Bray Curtiz

distance is also called Sorensen distance. Bray Curtiz distance

will have positive property if the compared values are also posi-

tive, and will have the values between 0 and 1:16

d�ab�=∑ �a−b�∑�a+b�

(8)

5. DESIGN EXPERIMENTFigure 1 showed the steps of this research. The first steps

is collecting the data from Region of Interest (ROI) Finger

Knuckle Print images with 440 samples from 55 individual to

be captured their images as many as 8 times. The image size

is 110 × 220 pixels, in form of grayscale. Based on 8 finger

3276


Fig. 1. Steps experiment.

knuckle print images of each individual, 5 images are taken to

be training images as many as 275 and 3 images are treated

as the Testing images as many as 165. The Database images

are gathered from Hong Kong Polytechnic University (http://

www4.comp.polyu.edu.hk/∼biometrics/FKP.htm).

Next step is pre-processing, we prepare the training and testing

data before feature extraction. This part, we resizes the image as

20×40 pixels to avoid out of memory while process recognition

in progress. In this phase, we conduct image quality improvement

using CLAHE in order to make the line feature inside the RoI

more clearly. In this step, the pre-processing without CLAHE

also conducted to compare the results with those using CLAHE.

Then, the extraction method is carried out to reduce the dimen-

sion using PCA in order to characterize the images based on the

image processing results. There are three type feature extraction

by PCA that have (60, 100 and 120) attribute.

Next, the measurement distance is conducted as recognition

method that compare and conform the train and test images by

calculating the differences between those two image vectors. The

last step is evaluating the distance measurement using highest

accuracy value using confusion matrix.

6. RESULTS AND ANALYSISThe results in the Table I, obtained from the pre-processing with-

out using CLAHE, in 60-feature PCA, the Euclidean Distance

showed accuracy until 90.30%, Manhattan distance as much as

89.09%, Minkowski distance as much as 87.27%, Chebyshev

distance as much as 80.60%, Chi square distance as much as

88.48%, Canberra distance as much as 76.36% and Bray Curtis

distance as much as 89.69%.

Table I. Accuracy of finger knuckle print without using CLAHE.

Distance measurement PCA 60 PCA 100 PCA 120

Euclidean distance 90.30 89.69 89.69

Manhattan distance 89.09 88.48 88.48Minkowski distance 87.27 87.87 87.87Chebychev distance 80.60 80.60 80.60Chi square distance 88.48 87.87 86.06Canberra distance 76.36 76.96 79.39Bray curtis distance 89.69 89.69 88.48

Table II. Accuracy of finger knuckle print using CLAHE.

Distance measurement PCA 60 PCA 100 PCA 120

Euclidean distance 80.60 80.60 81.21Manhattan distance 93.33 94.54 95.75

Minkowski distance 70.90 70.90 70.90Chebychev distance 64.84 64.84 64.84Chi square distance 95.15 95.15 95.75

Canberra distance 89.69 85.45 84.24Bray curtis distance 92.72 94.54 95.15

In 100-feature PCA, Euclidean distance, has accuracy until

89.69%, Manhattan distance as much as 88.48%, Minkowski

distance as much as 87.87%, Chebyshev distance as much as

80,60%, Chi square distance as much as 87.87%, Canberra dis-

tance as much as 76.96% dan Bray Curtis distance until 89.69%.

In 120-feature PCA, Euclidean distance has accuracy until

89.69%, Manhattan distance until 88.48%, Minkowski distance

as much as 87.87%, Chebyshev distance as much as 80.60%, Chi

square distance sebesar 86.06%, Canberra distance as much as

79.39% and Bray Curtis distance as much as 88.48%.

The results in the Table I, obtained from the pre-processing

without using CLAHE, show that by using 60-reduction, the

Euclidean Distance will have the accuracy value of 90.30%,

Manhattan distance as much as 89.09%, Minkowski distance as

much as 87.27%, Chebyshev distance as much as 80.60%, Chi

square distance as much as 88.48%, Canberra distance as much

as 76.36% and Bray Curtis distance as much as 89.69%.

In the Euclidean distance 100-reduction, it has the accu-

racy value of 89.69%, Manhattan distance as much as 88.48%,

Minkowski distance as much as 87.87%, Chebyshev distance

as much as 80,60%, Chi square distance as much as 87.87%,

Canberra distance as much as 76.96% dan Bray Curtis distance

sebesar 89.69%.

In the Euclidean distance 120-reduction, it has the accu-

racy value of 89.69%, Manhattan distance sebesar 88.48%,

Minkowski distance as much as 87.87%, Chebyshev distance as

much as 80.60%, Chi square distance sebesar 86.06%, Canberra

distance as much as 79.39% and Bray Curtis distance as much

as 88.48%.

Results in the Table II, obtained from the pre-processing using

CLAHE, show that by using 60-reduction, the Euclidean Dis-

tance will have the accuracy value of 80.60%, Manhattan distance

as much as 93.33%, Minkowski distance as much as 70.90%,

Chebyshev distance as much as 64.84%, Chi square distance as

88.68 87.67

Average of Accuracy FKP Recognition without Pre-processing CLAHE

87.47

77.57

89.29

80.60

89.8995.00

90.00

85.00

80.00

75.00

70.00

Euclide..

.

Manha..

.

Minko

...

Chebyc

...

Chi...

Canber

...

Bray..

.

Fig. 2. The accuracy average of FKP without CLAHE.

3277


Average of Accuracy FKP Recognition withPre-processing CLAHE

80.80

94.54

120.00

80.00

100.00

60.00

20.00

0.00

40.00

Euclid...

Manha..

.

Minko

...

Cheby..

.

Chi...

Canbe..

.

Bray..

.

70.90 64.84

95.35

86.46

94.14

Fig. 3. The accuracy average of FKP using CLAHE.

much as 95.15%, Canberra distance as much as 89.69% and Bray

Curtis distance as much as 92.72%.

In Euclidean Distance 100-reduction, it has the accuracy value

of 80.60%, Manhattan distance as much as 94.54%, Minkowski


64.84%, Chi square distance as much as 95.15%, Canberra dis-

tance as much as 85.45% and Bray Curtis distance as much as

94.54%.

In Euclidean Distance 120-reduction, it has the accuracy value

of 81.21%, Manhattan distance as much as 95.75%, Minkowski


64.84%, Chi square distance as much as 95.75%, Canberra dis-

tance as much as 84.24% and Bray Curtis distance as much as

95.15%.

In the other hands, accuracy of Euclidean, Minkowski and

Chebyshev distance are decreased when applied pre-processing

CLAHE. This conditions are showed the uses of pre-processing

CLAHE not always gives significant implication even quality of

images has been improved.

The experiment results in the Figure 2 are the values from the

accuracy average without CLAHE, showing that the Euclidean

distance has the accuracy value of 89.89%, Manhattan distance



much as 87.47%, Canberra distance as much as 77.57% and Bray

Curtis distance as much as 89.28%. Thus, in the pre-processing

step without using CLAHE, the highest accuracy is showed in

the distance measurement using Euclidean Distance as much as

89.89%.

Figure 3 shows the experiment results based on average values

from the pre-processing accuracy of Finger Knuckle Print using

CLAHE by using the distance measurement in the 60, 100 and

120 reductions.

The experiment results in the Figure 3 are the values from

the accuracy average using CLAHE, showing that the Euclidean

distance has the accuracy value of 80.80%, Manhattan distance



much as 95.35%, Canberra distance as much as 86,46% and Bray

Curtis distance as much as 94.14%. Thus, in the pre-processing

step without using CLAHE, the highest accuracy is showed in

the distance measurement using Euclidean Distance as much as

89.89%.

Figures 2 and 3 are calculate the total average accuracy with

feature extraction PCA 60, 100, and 120 in all distance measure-

ment. In Figure 2 have showed Euclidean as Best distance mea-

surements technique that has average accuracy until 89.89% in

recognition FKP without pre-processing. The experiment results

in the Figure 3 are showed Chi square distance measurement has

highest average accuracy until 95.35% with improvement accu-

racy until 8.2% compared with non-preprocessing.

7. CONCLUSIONThis experiment shows the use of different distance measurement

highly influences the accuracy of recognition. Moreover, the uses

of pre-processing CLAHE to improve image quality has shown

significant improvement in several distance measurements such

as Manhatan, Chi-Square, Canberra and Bray Curtis. However,

the implication of CLAHE in some distance gives worst result

that caused distance measurement such as Euclidean, Minkowski,

and Chebychev are decreased.

The used of Chi-Square distance are considered as best dis-

tance measurements that have best accuracy in recognizing

Finger Knuckle Print with pre-processing CLAHE that showed

accuracy until 95.35%.

References and Notes1. K. Usha and M. Ezhilarasan, Computers and Electrical Engineering

(2014).2. L. Zhang, L. Zhang, D. Zhang, and H. Zhu, Pattern Recognition 43, 2560

(2010).3. Z. S. Shariatmadar and K. Faez, Optik—International Journal for Light and

Electron Optics 125, 908 (2014).4. S. Ribaric and I. A. Fratric, IEEE Trans. Pattern Anal. Mach. Intell. 27, 1698

(2005).5. A. K. Jain, A. Ross, and S. Pankanti, A prototype hand geometry based

verification system, Proceeding AVBPA, Washington, DC (1999), Vol. 1,pp. 166–71.

6. Z. Lin, Z. Lei, and Z. David, Finger-Knuckle-print: A new biometric identifier,Proceeding IEEE International Conference on Image Processing, Cairo, Egypt(2009), Vol. 1, pp. 76–82.

7. D. L. Woodard and P. L. Flynn, Comput. Vis. Image Underst. 100, 357 (2005).8. N. Ozkaya and N. Kurat, Journal of Visual Communication and Image Repre-

sentation 25, 1647 (2014).9. Z. Lin, Z. Lei, and Z. David, Finger-Knuckle-Print Verification Based on

Band-Limited Phase-Only Correlation, LNCS 5702, Springer-Verlag, Berlin,Heidelberg (2009), Vol. 1, pp. 141–8.

10. Y. Pengfei, Z. Hao, and L. H. Yan, Appl. Mech. Mater. 44, 703 (2014).11. C. Ravikanth and A. Kumar, Biometric authentication using finger-back sur-

face, Proceedings of the CVPR’07 (2007), pp. 1–6.12. A. Kumar and C. Ravikanth, IEEE Transactions. Information Forensics and

Security 4, 98 (2009).13. G. Suprijanto, J. E. Azhari, and L. Epsilawati, Image contrast enhancement for

film-based dental panoramic radiography, International Conference on SystemEngineering and Technology (2012).

14. S. M. Pizer, E. P. Amburn, J. D. Austin, R. Cromartie, A. Geselowitz, T. Greer,B. T. H. Romeny, J. B. Zimmerman, and K. Zuiderveld, Computer VisionGraphics and Image Processing 39, 355 (1987).

15. M. Turk and A. Pentland, J. Cognitive Neuroscience 3, 71 (1991).16. S. Szabolcs, Color histogram features based image classification in content-

based image retrieval systems, 6th International Symposium on AppliedMachine Intelligence and Informatics (2008), pp. 221–224.

17. D. C. Adams, F. J. Rohlf, and E. S. Dennis, Italian Journal of Zoology 71, 5(2004).

18. V. Asha, GLCM Based Chi-Square Histogram Distance for Automatic Detec-tion of Defects on Patterned Textures.


3278




Vol. 21, 3289–3292, 2015

Distortion Analysis of Indoor and Outdoor Limit

with Biconical Antenna for Ultra Wideband System

Chairak Deepunya∗ and Sathaporn Promwong

Faculty of Engineering, King Mongkut’s Institute of Technology Ladkrabang, Bangkok 10520, Thailand

The distortion of ultra wideband impulse radio (UWBIR) system that distorted by a channel due to antennadispersion. This highly degrades of link budget performance. Therefore, to know the antenna characteristics,the effects of a waveform distortion, are necessary. The UWB transmission waveform is investigated by usingextended Friis transmission formula based on UWB measurement data. This paper evaluates the performancesystem on transmission waveform distortion for UWB communication. The received signal and isotropic signaltemplates are considered with Friis formula. Mostly of radio wave propagation in link budget analytic performanceis evaluated by using the Friis equation. Due to Friis equation is cannot directly apply for the UWB transmis-sion waveform. For experimental evaluation scheme, using the broadband antenna for transmitter and receiverantennas (Tx and Rx). The full band 3.1 GHz to 10.6 GHz, which contributed by the Federal CommunicationsCommission (FCC), is proposed as the UWB signal waveform. The transfer functions measured as experimentalresult by using the vector network analyzer for measuring and recording. This paper UWB transmission gainswith the received signal and isotropic signal templates are shown and compared.

Keywords: UWB, UWB Antenna, Link Budget, Distortion.

1. INTRODUCTIONThe nowadays UWB system is new for wireless system has

become important for short range wireless communications sys-

tem with its high speed communication, low power consumption

potentials and low.1�2 Therefore, the technology of UWB is dif-

ferent from other technologies. The UWB radio transmits is with

impulse waveform and wider bandwidth of the norrowband spec-

trum. The FCC regulation provides that UWB spectrum from

3.1 GHz–10.6 GHz,13 and also has been greater than 0.20 of the

fractional bandwidth. The UWB fractional bandwidth defined as

BWf = 2�fH −fL�

fH −fL≥ 0�20 (1)

where fL is minimum and fH is maximum of frequency.

UWBIR transmission waveform under the FCC, with part

15 limits has power spectral density under −41.3 dBm/MHz,

which taken as a lower noise floor. Thus, the reason why

UWB communication system can coexist within other microwave

communications and propagation engineering. Furthermore, the

UWBIR system is an ideal trend that new communication tech-

nology for wireless communication, short impulse radio system,

high speed transmission rate and low cost technology for indoor

systems in covered wireless personal area networks (WPAN) and

wireless body area network (WBAN).3


In the communication systems has used Friis transmission for-

mula and it is widely used for calculating the propagation chan-

nel for narrowband communications.4 The complex form in the

free space of Friis formula expression is modified for UWBIR

system.5–7 The UWBIR system, conceded used matched filter.8–10

Although, the frequency spectrum and signal distorted by energy

channel transfer function is used for deriving the SNR gains,11

if considerations about the measured frequency transfer function

and UWB antenna transfer function.

The performance evaluation of waveform distortion due to

channels and antennas for UWB impulse radio. The UWB

receiver considering using received signal and isotropic signal

templates provided to analyze of noise level between input wave-

form and output waveform is obtained. Herein, both biconical

antennas are considered both at transmitter (Tx) and receiver

(Rx) sides. Full band spectrum frequency, which is followed FCC

both indoor and outdoor spectral mask, provided as the UWB

transmitted.12�13 The UWB transmitted is measured and recorded

by using a VNA work done in an anechoic chamber. The band-

width of measurement is from 3–11 GHz. The waveform distor-

tions are considered for the UWBIR transmission waveform.

The transmission gain model of UWB impulse radio receiver

with received signal and isotropic signal templates is shown and

compared with the experimental results and conclusion. This

scheme provides some useful physical insights and optimized

design procedure with a clear and accessible description of the

UWB link budget comprised of practical antennas.



In this organized of paper as follows. Section 2 the UWB

impulse radio measurement system. Section 3 the distortion anal-

ysis of UWB impulse radio system. Next, the experimental

results and discussion are illustrated in Section 4. Finally is con-

clusion in section 5.

2. UWB IMPULSE RADIO MEASUREMENT

SYSTEM2.1. UWB Waveform Model

The transmission waveform is more obvious in the UWBIR sys-

tem. In this paper, the transmitted waveforms that fully following

FCC from 3.1 GHz to 10.6 GHz13 and common frequency band

FCC in USA are considered. Also, the root raised cosine (RRC)

model is used as the waveform transmitted.

The waveform as passband in root raised cosine is written as

Vt�ro�f �=

⎧⎪⎪⎪⎪⎪⎨⎪⎪⎪⎪⎪⎩

1 �f �−fc � ≤�1−��

2T

A�1−��

2T< ��f �−fc � ≤ �1+��

2T

0 otherwise

(2)

where

A=√

1

2

[1+cos

(�T

�

[��f �−fc �−

1−�

2T

])](3)

T = 1/fb denotes the reciprocal of the symbol-rate and fb is

the spectral bandwidth, fc denotes the center frequency, �= 0�3satisfies roll-off factor. For following FCC spectral masks, fc was

6.85 GHz. The spectral bandwidth fb was 6.37 GHz. For follow-

ing the common frequency band spectral mask, then fc and fbwere 7.877 and 0.975 GHz respectively.

The power spectral density normalized of these waveforms sat-

isfying FCC spectral masks shown in Figure 1.

2.2. Experimental System

The UWB waveform distortion is experimented are considered

used VNA for measuring and recoding in free space. The VNA

Fig. 1. The power spectral density with root raised cosine pass band wave-form by satisfying FCC spectral mask.13

operates in the channel sounding mode, where Two-ports are Tx

antenna and Rx antenna respectively. Both Tx and Rx antennas

height at 1.75 m and separated by 4 m. This experimental setup

is shown in Figure 2 and antenna orientation was rotated Rx

antenna rotation from 0� to 360� by step 5� and antenna polar-

ization in horizontal is measured.

In the calibration techniques of UWB transmission link was

conducted by back to back transmission and short open load

thru (SOLT). Therefore, all impairments of characteristics of the

biconical antennas are verified with the measured results.

3. DISTORTION ANALYSIS OF UWB SYSTEM3.1. Transmission Waveform

The UWB transmission waveform, free space link budget is

transformed by a transfer function. Thus, in free space transfer

function Hf �f � is written as

Hf �f �d� = c

4�fde−j2�fd/c (3)

The transfer function in free space Hc�f � with the biconical

antennas is obtained based on modified of the channel as

Hc�f �=Hf �f �d�Ht�f ��t� ·Hr�f ��r� (4)

where Hs�f ��s� (s = n or m) represents a complex of channel

between transmitted antenna and received antenna with differ-

ence polarization the �s = ��s��s� as

Hs�f ��s� = Hs�f ��s��s�

= �sHs�f ��s��s + �sHs��f ��s��s� (5)

which has the relation as

1

4�

∫ 2�

0

∫ �

0

�Hs�f ��s��s��2 sin� d� d�= �s (6)

where �s denotes the antenna efficiency, in addition, the solution

can be normalized with isotropic antenna.

3.2. UWB Correlation Receiver

The UWB correlation receiver is considered by using template

waveform, which is shown in Figure 3. The template received

signal and isotropic received signal are analyzed based on input

and output conrrelation waveforms are considered. In this paper,

Fig. 2. The experimental setup.

3290


Fig. 3. System modeling of UWB system.7

the optimal receiver is proposed from optimum signal with

received signal of UWB channel.

In this section the condition of noise channel between are

discussed, the transmission channel of receiving signals and

isotropic signal templates, Ho�f � and Hi�f � are prospectively

normalized as∫ �

−��Ho�f ��2 df =

∫ �

−��Hi�f ��2 df = 2fb (7)

In addition, the condition of noise bandwidth is formulated

by N0fb , where N0/2 represented additive white Gaussian noise

(AWGN).

The UWB optimum receiver with received signal and isotropic

signal templates is written as

Hopt�f �=√

2fbV∗r �f �√∫�

−� �Vr�f ��2 df(8)

Hi�f �=√

2fbV∗r−i�f �√∫�

−� �Vr−i�f ��2 df(9)

where �·�∗ the free space complex conjugate, Vr�f � and Vr−i�f �denote the spectrum frequency of receiving signals of channel

measured and isotropic antennas respectively. Then, the expres-

sion from can be shown as

Vr�f �=Hc�f �Vt�f � (10)

Vr−i�f �=Hf �f �Vt�f � (11)

where Vt�f � represents the spectrum waveform as related with

Fourier transform

Vt�f �=∫ �

−�vt�t�e

−j2�f t dt (12)

Hc�f � represents the channel vector as from Section 3.1 and

Hf �f � is, which given by

Hf �f �d� =c

4�d�f � e−j2�fd/c (13)

where d denotes distance of between transmit and receive

antenna and c is the velocity of light.

3.3. Transmission Gains

The UWB signal distortion to evaluate the peak value of the cor-

relation receiver output of biconical antennas simplified that the

received signal and isotropic signal templates are compared. The

waveform distortion from antenna is normalized with correlation

template receiver, the UWB transmission gain is represented in

the UWB transmission gain of the signal-to-noise ratio (SNR) at

the UWB impulse radio.

Fig. 4. UWB measurement gain of rectangular by using FCC underbiconical–biconical antennas transmission link.

Therefore, UWBIR transmission of correlation receiver as tem-

plate waveform GWM is written by

GWM =max

∣∣∣∣∫�−� vr �t�hWM�t−��dt

∣∣∣∣max

∣∣∣∣∫�−� vr−iso�t�hWC�t−��dt

∣∣∣∣(14)

The gain of UWBIR transmission and the isotropic template,

where GWC as

GWC =max

∣∣∣∣∫�−� vr �t�hWC�t−��dt

∣∣∣∣max

∣∣∣∣∫�−� vr−iso�t�hWC�t−��dt

∣∣∣∣(15)

Fig. 5. UWB measurement gain of RRC by using indoor FCC underbiconical–biconical antennas transmission link.

3291


Fig. 6. UWB measurement gain of RRC by using outdoor FCC underbiconical–biconical antenna transmission link.

4. RESULTS AND DISCUSSIONThe evaluation of waveform distortion in the UWBIR transmis-

sion model are considered of transmission gain are shown. First,

the UWB transmission gain of receiving a signal of receiving

signals and isotropic signal templates at the receiver are consid-

ered. The UWB measurement gain of two passband waveform is

shown. In particular of FCC, the rectangular passband, the RRC

passband verifying indoor scenario and outdoor surrounding are

in Figures 4 to 6, respectively. The comparison of experimen-

tal results is compared to receive signals and isotropic templates

following FCC, which can know clearly that using a common

frequency band is better than FCC. At the FCC, the estimated

resultant of the rectangular passband, the RRC passband com-

pare both indoor scenario and outdoor surrounding as 1.21, 1.32

and 1.27 dB as completely.

5. CONCLUSIONIn this paper, the biconical antennas are investigated as a broad-

band antenna to educate for UWB communication system, also

waveform distortion due to channels and antenna measurement

has been proposed. The waveform distortion of channel transfer

function and biconical antennas are evaluated for UWB impulse

radio system by using extended Friis transmission formula. The

correlation receiver with received signal and isotropic signal tem-

plates are evaluated. Using a biconical antenna is applied as the

transmit and receive antennas. As the results, the relative gains in

the received signal and isotropic signal templates are every small

difference.

References and Notes1. G. Adamiuk, UWB antenna for communication systems, Proceeding of the

IEEE, April (2012), Vol. 100, pp. 2308–2321.2. V. Yajnanarayana, Design of impulse radio UWB transmitter for short

range communication using PPM signals, IEEE International Conference onElectronics, Computing and Communication Technologies, January (2013),pp. 1–4.

3. J. Farserotu, A. Hutter, F. Platbrood, J. Gerrits, and A. Pollini, Wireless Per-sonal Communications 197 (2002).

4. H. T. Friis, A note on a simple transmission formula, Proc. IRE, May (1946),Vol. 34, pp. 254–256.

5. J. Takada, S. Promwong, and W. Hachitani, Extension of Friis’ transmissionformula for UWB systems, Technical Report of IEICE, WBS2003-8/MW2003-20, May (2003).

6. S. Promwong, W. Hachitani, and J. Takada, Experimental evaloation schemeof UWB antenna performance, Technical Meeting on Instrument and Mea-surement, IEE Japan, IM-03-35, June (2003).

7. S. Promwong, W. Hachitani, J. Takada, P. Supanakoon, and P. Tangtisanon,Experimental study of ultra-wideband transmission based on Friis’ transmis-sion formula, The Third International Symposium on Communications andInformation Technology (ISCIT) 2003, September (2003), Vol. 1, pp. 467–470.

8. S. Promwong, J. Takada, P. Supanakoon, and P. Tangtisanon, Path lossand matched filter gain for UWB system, 2004 International Symposium onAntenna and Propagation (ISAP), August (2004), pp. 97–100.

9. S. Promwong, J. Takada, P. Supanakoon, and P. Tangtisanon, Path loss andmatched filter gain of free space and ground reflection channels for UWB radiosystems, IEEE TENCON 2004 on Analog and Digital Techniques in ElectricalEngineering, November (2004), pp. 125–128.

10. F. Tufvesson and A. F. Molisch, Ultra-wideband communication using hybridmatched filter correlation receivers, 2004 IEEE 59th Vehicular TechnologyConference (VTC), May (2004), Vol. 3, pp. 1290–1294.

11. P. Supanakoon, K. Teplee, S. Promwong, S. Keawmechai, and J. Takada,Theoretical SNR gain and BER performances of UWB communications withmatched filter and correlation receivers, The International Technical Con-ference on Circuits/Systems, Computers and Communications (ITC-CSCC2006), July (2006) pp. 269–272

12. P. Supanakoon, K. Wansiang, S. Promwong, and J. Takada, Simple waveformfor UWB communication, The 2005 Electrical Engineering/Electronics, Com-puter, Telecommunication, and Information Technology International Confer-ence (ECTI-CON 2005), May (2005), pp. 626–629.

13. Federal Communications Commission, Revision of Part 15 of the Commis-sion’s Rules Regarding UWB Transmission Systems, First Report, FCC 02-48,April (2002).

14. W. Hirt and M. Weisenhorn, Overview and implications of the emerging globalUWB radio regulatory framework, Proceeding the 2006 IEEE InternationalConferences on Ultra-Wideband, September (2006), pp. 581–586.

15. J. Foester, Channel Modeling Sub-Committee Report Final, IEEEP802.15-02/368r5-SG3a, November (2002).


3292




Vol. 21, 3347–3351, 2015

Big Data, Cloud and Bring Your Own Device:

How the Data Protection Law Addresses the

Impact of “Datafication”

Sonny Zulhuda1�∗, Ida Madieha Abdul Ghani Azmi1, and Nashrul Hakiem2

1Civil Law Department, International Islamic University Malaysia2Faculty of Science and Technology, UIN Syarif Hidayatullah Jakarta

Many hypes are currently surrounding the “datafication” such as the Big Data, Cloud and BYOD. The proliferationof data from ubiquitous sources is often not counter-balanced with adequate awareness and prudent risk man-agement by the end-users, making it easier for others to take advantage of the new technology and reap fromthe abundant data available for all kinds of purposes including criminal. IT stakeholders should view Big Datanot only as a new exciting technological advancement, but also a frontier full of potential risks to be addressednot only by industrial best practices, but also by the reforming laws in the area of information security and dataprivacy. This paper sets to undertake two major tasks. Firstly, examining the types of legal risks involved in theBig Data environment. Secondly, it highlights some aspects of data privacy and security laws already containedin the current data protection laws in Malaysia.

Keywords: Datafication, Big Data, Data Security, Personal Data Protection, Malaysia.

1. INTRODUCTION“Datafication” is, in the words of Meyer-Schonberger and Cukier,

the process of quantifying all information around us: our loca-

tion, movement, communications, usage of devices, etc. which

will allow us to use such information in new ways, such as in pre-

dictive analysis. This will help us further to unlock the implicit,

latent value of the information.1 The amount of those quanti-

fied data may prove to be unimaginable, hence the expression

“Big Data.”

“Big Data” is an expression that typically refers not only to

specific, large datasets, but also to data collections that consoli-

date many datasets from multiple sources, and even to the tech-

niques used to manage and analyze the data.2 According to the

report by a US official report, Big Data is big in two differ-

ent senses. It is big in the quantity and variety of data that are

available to be processed. And, it is big in the scale of analysis

(“analytics”) that can be applied to those data, ultimately to make

inferences.3

It is an ocean of data out there where we can only swim or

sink.4 This datafication is poised to reshape the way we live,

work and think.1 Those data are abundant. It comes in a mas-

sive volume, velocity and variety that are unprecedented.5 In the

words used by the Gartner IT glossary, Big Data is defined as

“high-volume, high-velocity and high-variety information assets


that demand cost-effective, innovative forms of information pro-

cessing for enhanced insight and decision making.”6

Another immediate effect of datafication is the emergence of

“cloud computing.” The Big Data environment has in turn led

the companies to tap server capacity as needed to accommodate

an enormous scale required to process big datasets and run com-

plicated mathematical models. Here is the nexus between Big

Data and cloud computing.7 The massive dataset that an organ-

isation is having at its disposal presents another challenge: how

can it store, process and exploit such big data in the best and

most efficient manner? This is what makes cloud computing an

ultimate choice. Being hailed as the future of information tech-

nology (IT) architecture, by the year 2018 cloud computing is

projected to be a major medium for delivery of information and

other IT functions at both the consumer and corporate ranks.8

The US National Institute of Standards and Technology (NIST)

defines cloud computing as “a model for enabling ubiquitous,

convenient, on-demand network access to a shared pool of con-

figurable computing resources. That can be rapidly provisioned

and released with minimal management effort or service provider

interaction.”9 Hoover reckoned that the businesses increasingly

see cloud computing as a “valuable proposition for decreas-

ing technology costs, enabling and accelerating the delivery of

new technology services, and refocusing technology workers on

mission-oriented tasks that deliver more business value than time

spent maintaining corporate technology systems.”10



Nevertheless, cloud computing is not a risk-free concept as

it also has a tremendous implication on what sets of new risks

an organisation will encounter in their data management. This

is because the main issue with a cloud computing is that the

owner or controller of the data has no longer assumed a complete

and exclusive control over their data assets: some portions of

the analytics and processing are being outsourced to an external

entity.

Meanwhile, this Big Data environment compels companies to

be become increasingly agile, productive and focused in-order

to achieve a competitive advantage in the market place. This

development pushes for renewed working options such as resort-

ing to cloud computing and IT service outsourcing. Besides that,

to create more fluidity and flexibility at work, employers allow

staffs to work virtually from any place but the office itself. They

typically tell the employers that they can bring their own work

device to and from home—hence the “Bring Your Own Device”

(“BYOD”) concept.

This paper is making an attempt to revisit this aspect, namely

the risks of datafication to an organisation, more specifically on

the practices of cloud computing and BYOD and to analyse it

from the legal perspective. It employes a legal doctrinal research

that collects primary data from the statutory/legislative provi-

sions and secondary data from case reports. It observes cases

where data resources management was found problematic, then

explores the relevant legal and regulatory requirements on the

area of personal data protection retrieved from the Malaysian

legal framework.

The contribution brought about by this paper is in its strategic

objective: it seeks to prove how the big data environment today

gives rise to intersection with data protection legal requirements.

It highlights the need for a future research on organisational

road-map to comply with data protection requirements in more

structured and systematic manner as part of organisation’s data

governance.

2. CHALLENGES OF PERSONAL

DATA MISMANAGEMENTGiven the promises of datafication, companies and individuals

are quickly trying to grab every possible fortune that pops up

in this process. However the process of embracing the Big Data

environment is often marked with its dark side. The individu-

als whose personal information have increasingly flooded public

spaces and open network are normally the immediate receiver of

the consequences. Their data are thus exploited: discreetly stored,

aggregated, disclosed, exchanged for money, traded at black mar-

ket or kept indefinitely. Thus it is natural that each stakeholders

of this Big Data are expected to manage their data well so as to

avoid those risks.11

The dark side to Big Data takes many forms, such as an

interference with privacy.12 For example, the extensive amounts

of personal information revealed during online transaction have

taken the relationship between customer profiling, predicting

trends, and marketing to a whole different level. As Reyhaneh

explains, Big Data is capable of tracking movements, behaviour

and preferences, and predicting the behaviour of individuals with

unprecedented accuracy. The more access business has to Big

Data, the better it can target us with advertising that matches

(or predicts) our specific interests. This is, however, often done

without our consent.12

As for the cloud computing, risk appears as early as it starts.

The transfers of organisations’ databases to external and central-

ized data centers means the transfer of certain duty of safeguard-

ing the data itself. Doubt would naturally grow when the security

measures are taken over by cloud providers, the latter may not

be fully trustworthy.8 Thus, with the explosion of data being

outsourced to this external parties, the increasing prevalence of

identity theft and data security breaches for cloud consumers is

paramount.8

Understanding the legal requirement on data security and pri-

vacy would be easier if one knows the evils and mischief of

data mismanagement. In the perspective of information gover-

nance and information security, personal data is an asset because

it has value. Outsiders would be interested to have access to its

customer database including external marketers, business com-

petitors or any adventurous individuals who grab any opportunity

to make use of such company’s database.

One can borrow cases from the UK’s Information Commis-

sioner Office to show data user’s mishandling of their personal

data resources in the following paragraphs.13

• It was reported in 2008 that the UK Financial Services Author-

ity (FSA) found three units under HSBC group had failed to

put in place adequate systems and controls to protect customers’

details from being lost or stolen. HSBC Life, HSBC Actuaries

and HSBC Insurance Brokers were fined a collective amount of

over 3 million pounds for having lost unencrypted disks contain-

ing personal details of their customers.

• In 2011, an encrypted memory stick containing patients’

sensitive personal data of the Arthur House Dental Care was

accidentally lost from the possession of one of the dentist. The

memory stick was used as a temporary back-up solution and was

taken home by a dentist for safekeeping. The memory stick was

found in public and was sent to the Information Commissioner

Office (ICO), who in turn took action against the dental care for

their failure to place proper safeguards in providing data backup

system.

• An e-mail containing a medical report of patient’s health was

mistakenly sent by Aneurin Bevan Health Board (ABHB) to a

wrong patient in Wales. The UK ICO found that there was not

enough robust system to prevent this case of accidental disclo-

sure. As part of enforcement notice, the data user had to imple-

ment some measures including staff data protection training,

monitoring of compliance, and the introduction of new checking

processes before personal information is sent out.

• A contractor company who was tasked by Brighton and Sus-

sex University Hospital to remove and destroy old computer

hard drives containing sensitive personal data was found to be

in breach of Data Protection Act after one individual from the

contractor company took some of the hard drives and sold them

on eBay. The ICO has levied a fine of £325�000 on BSUH over

the breach for failure to put proper supervision on the process of

data removal and on the involvement of a third party contractor.

The cases shown above, dealt with under the UK Data Pro-

tection Act 1998 should serve as an alarm to all data users. The

emerging legal rule under the new Act places a higher standard of

due diligence to be complied by companies. Previously, a com-

pany whose data was found to have unintentionally been leaked,

lost, or subjected to disclosure would only be potentially liable

if there is an evidence of negligence in their part, which likely

means establishing positive breach of duty of care. With new

3348


written laws emerging in many jurisdictions, such standards had

become more certain and measurable, some of which is portrayed

in this paper.

3. DATA PROTECTION LAW IN MALAYSIAMajor legal issues on data privacy in Malaysia were addressed

by the Personal Data Protection Act (PDPA) 2010. Being the

main legal framework for protecting data privacy of individuals,

PDPA regulates the processing of personal data in commercial

transactions.14 Under Section 4, “personal data” refers to any

“data that relates directly or indirectly to a data subject, who

is identified or identifiable from that information or from that

and other information in the possession of a data user, including

any sensitive personal data and expression of opinion about the

data subject.” Meanwhile, “commercial transactions” mean “any

transaction of a commercial nature, whether contractual or not,

which includes any matters relating to the supply or exchange of

goods or services, agency, investments, financing, banking and

insurance.”

The enactment of the PDPA is arguably a milestone for the

development of e-commerce in Malaysia, considering that a mas-

sive and increasingly valuable amount of personal information

are being stored, processed and exploited.14 At the heart of PDPA

is a set of duties under the data protection principles from which

stemming all the rights, duties and liabilities of each of data user

and data subject (“data user” is those who use, collect, process,

etc. the personal data that belong to certain individuals, i.e., the

“data subject”). There are generally seven categories of the duty

spelled out in Table I, whereas each of those principles con-

tributes to the data protection in Malaysia.

The most relevant provision under the Act would be the

offence of unlawful collecting or disclosing of personal data

(section 130). If any person is found to have knowingly or reck-

lessly collected or disclosed personal data that is held by the

data user without the consent of the latter commits an offence

punishable with a fine of maximum MYR500, 000.00 or with

imprisonment for a maximum term of three years or with both.

The same penalties await those who sell personal data under the

circumstances set out above. This provision does not specify the

manner of such collection, disclosure or selling of the personal

data but instead leaves it open so as to be able to catch offenders

Table I. PDP principles under the PDPA 2010.

Section Principle Information security implications

6 General principle No process of personal data which isexcessive and/or without the consent ofdata subject.

7 Notice and choice Proper notification on the purpose of thatdata collection/processing

8 Disclosure Prohibits unauthorized disclosure orsharing of personal data

9 Security Imposes security measures by data usersthat commensurate the risk of securitybreach

10 Retention Personal data shall not be keptunnecessarily

11 Data integrity Right of data subjects to correct andupdate their personal data

12 Data access Right of data subject to have an access tohis own personal data the at the user’sdatabase

in various ways or modus operandi. The provision, it is argued is

useful in providing strong and effective measures against unlaw-

ful collection and disclosure of personal data.

Another important provision is the duty of data users such as

those cloud service providers to conduct due diligence as to the

reliability and security of their electronic system. This is because

under section 133, the board of director or any officer responsible

for the management of company may be charged for an offence

by body corporate, unless he can prove his absence of knowledge,

and that he had taken all reasonable precautions and exercised

due diligence to prevent the commission of the offence. Given

this analysis, it can be said that the PDPA can lend a hand for

the security of data in the electronic environment.

4. CLOUD AND DATA PROTECTIONCloud computing can be understood as a way of delivering

computing resources as a utility service via network, typically

the Internet, scalable up and down to user requirements.15 The

major feature of cloud computing is that it enables decentral-

ization of resources such as data and applications system, thus

would enable an entity (e.g., companies) to free their resources

or at least be more focused on the functions and less on man-

aging their data resources. As a result, the data owner’s control

over their own data would substantially decrease while surrender-

ing more control to a third party to manage their data resources.

As such, a new risk is just getting created, namely the increasing

risk of data security and confidentiality breach.15

From data protection perspective, the core challenge in cloud

computing and IT service out sourcing would be the fact that data

user loses a complete control over personal data of their data sub-

jects to whom they are initially answerable. Huge amount of per-

sonal data will be processed by a third party on their behalf. With

the absence of immediate supervision and the physical bound-

aries, how can data users ensure the security of personal data?

In Malaysia, the legislature has stepped forward to put cer-

tain legal requirements that will change the landscape of cloud

computing services.14 Under the PDPA, those third parties who

provide cloud computing or other IT services are called “data

processor” simply because they process the personal data solely

on behalf of the data user, and do not process the data for any

of their own purposes. The moment they process it for their own

purpose, they become a “data user” for that particular purpose.

The following is a set of duties and responsibilities of a data

processor that can be derived or extracted from the provisions of

the PDP Act 2010:

(a) The security principle in section 9 mandates that data user

shall take measures for ensuring the reliability, integrity and

competence of personnel having access to the personal data.

This in turn requires the establishment of an agreed and well-

defined roles and responsibilities between data processor and

data user in relation to the processing of personal data. This can

be achieved with a well-crafted PDPA-compliant service level

agreement (SLA);

(b) Following the above based on the same provision, it is there-

fore desirable that data users and data processors have in place

internal policies, standards or procedures of information security

measures that govern the processing of data (including that of

protective, detective and responsive measures). These measures

include technical and organizational ones.

3349


(c) Based on section 133 on the offence by body corporate that

has been mentioned earlier, the data users and data processors

will have to implement due diligence to ensure and maintain

compliance with the contractual guarantees and internal mea-

sures above. This “due diligence” may be generally interpreted as

putting in place a proper system of data management and proper

monitoring of such system (See: Universal Telecasters (Qld) Ltd.v Guthrie [1978] FCA 18);

(d) By virtue of section 101 of the PDPA 2010, data users and

data processors are both subject to inspection and compliance

audit by Commissioner to the information system used for the

processing of personal data.

(e) On top of that, there are more stringent rules when it involves

a foreign party as cloud provider. The PDPA requires a two-

layer safeguard including authority clearance and contractual

safeguards. This is provided in section 129 of the PDPA 2010.

Based on the above, it is important to reiterate here that in the

first place it is the duty of every data user to ensure that such

PDPA-compliant agreement takes place between them and data

processors. The obligations of data processors such as the cloud

service providers will necessarily reflect those duties of the data

users, as long as the cloud providers process the data on behalf

of the data users, and not for their own purpose.

5. BYOD–“BRING YOUR OWN DEVICE”Another recurring challenge with personal data management is

about the trend of “bringing your own device” or BYOD. It is

a commendable development that employees are more mobile

and even free to bring their own device to work from home or

anywhere else but their office. However, BYOD can be risky if

one does not execute proper threat analysis and risk management

measures. The above cases of personal data breaches due to the

lost or stolen devices should serve as lessons for all of us.

Under the security principle, the PDPA clearly mandates that

data user shall, when processing personal data, take practical

steps to protect the personal data from any loss, misuse, modifi-

cation, unauthorized or accidental access or disclosure, alteration

or destruction. By virtue of section 9 of PDPA 2010, such prac-

tical steps should take into account:

• The nature of the personal data and the harm that would result

from such loss, misuse, modification, unauthorized or accidental

access or disclosure, alteration or destruction;

• The location where the personal data is stored;

• Any security measures incorporated into any equipment in

which the personal data is stored;

• The measures taken for ensuring the reliability, integrity and

competence of personnel having access to the personal data; and

• The measures taken for ensuring the secure transfer of the

personal data.

Given the above, the authors outline the following as necessary

safeguards to ensure BYOD works well for the organisations in

big data environment:

(a) Identity data asset classification. This means data users

should be aware of the nature of the harm on each of class of

personal data in their possession. Only then they can determine

which personal data should be kept at which security level.

(b) Ensure and manage proper access control. In other words,

data users should identify who can bring home what device and

what data. Data users should therefore apply the principle of

least privileged.

(c) Distribution of role between employees. This is important

because not all employees need to bring their devices home.

(d) Risk management, business continuity plan and disaster

recovery management. This is because the Act requires data users

to carefully determine the right measures taken for ensuring the

secure transfer of the personal data.

(e) Due care and diligence: On top of what has been mentioned

on similar issue in the earlier passage, it is noteworthy that the

above measures are collective responsibility, not only that of the

IT division in an organization. The precaution and due diligence

to be taken by all members in an organisation is to encourages

a massive overhaul in redefining each role and responsibility

required in the data processing activities.

6. CONCLUSIONGiven the analysis above, we can conclude that the data protec-

tion law has been developed in major parts to respond to the chal-

lenges of datafication and trends associated with it. In Malaysia,

this attempt is arguably spearheaded by the Personal Data Pro-

tection Commissioner established under the PDP Act 2010. The

personal data law mandates certain steps of legal risks man-

agement to be taken by business entities in securing their data

resources. Compliance with the law helps organizations in safe-

guarding their assets in this Big Data environment.

Personal data law is not only concerned with protecting “per-

sonal” interest such as the privacy of individuals, but it is also

about safeguarding organizational assets in the whole lifecycle

of such data. Such law oversees not only on the lifecycle of

organizational data asset, but also the whole processes of IT

governance involving people, process and technology. Further-

more, it is not only about data management, but also personnel

awareness, robust system protection, and good business and gov-

ernance (such as the requirement of data due diligence). Com-

pliance with data privacy law is arguably a passport to a global

business because personal data protection is indeed becoming

another widely-accepted indicator of transparency and good gov-

ernance and can be potential trade barrier in the future.

This paper reiterates that today’s business landscape has

changed. With data increasingly becomes bigger and faster, risks

and threats need to be more comprehensively managed. Technol-

ogy alone is not an answer. It has to be further accompanied by a

paradigm shift in the mindset of people and processes involved.

The more vigilant one is, the more he can control the risks and

challenges posed by the big data environment.

It is reiterated here that this paper undertakes the task to estab-

lish the link between requirements of protecting personal data

as mandated by the data protection law with the contemporary

trends in big data environment today. It also paves the way for-

ward to highlight the need for a future research on organisational

road-map to comply with data protection requirements in more

structured and systematic manner as part of organisation’s data

governance.

References and Notes1. V. Mayer-Schonberger and K. Cukier, Big Data: A Revolution That Will Trans-

form How We Live, Work and Think, John Murray, London (2013).2. M. R. Wigan and R. Clarke, IEEE Computer 46 (2013).

3350


3. Big Data and Privacy: A Technological Perspective, President’s Council ofAdvisors on Science and Technology. May 2014. Available at: https://www.whitehouse.gov/sites/default/files/microsites/ostp/PCAST/pcast_big_data_and_privacy_-_may_2014.pdf. Accessed April 30, (2015).

4. Intel, What Happens in an Internet Minute?, 2014, Available at: http://www.intel.com/content/www/us/en/communications/internet-minute-infographic.html. Accessed April 30 (2015).

5. W. H. Davidow, Overconnected: The Promise and Threat of the Internet, Del-phinium Books, New York (2012).

6. ICO. Big data and data protection. 20140728. Version: 1.0. Available at:https://ico.org.uk/media/for-organisations/documents/1541/big-data-and-data-protection.pdf. Accessed April 30 (2015).

7. HM GovernmentHorizon Scanning Programme, 2014, Emerging Tech-nologies: Big Data, TheEmerging Technologies Big Data Community ofInterest. December 2014. Available at: https://www.gov.uk/government/

uploads/system/uploads/attachment_data/file/389095/Horizon_Scanning_-_Emerging_Technologies_Big_Data_report_1.pdf. Accessed April 30 (2015).

8. J. A. Harshbarger, Journal of Technology Law and Policy (J. Tech. L. andPol’y) 16, 229 (2011).

9. J. Ryan, Santa Clara Law Review 54, 497 (2014)

10. J. N. Hoover, Journal of Business and Technology Law (J. Bus. and Tech. L.)8, 255 (2012).

11. N. Gifford, Information Security–Managing the Legal Risks, CCH AustraliaLtd., Sydney (2009).

12. R. Saadati and A. Christie Internet Law Bulletin 67 (2013).13. ICO. UK’s Information Commissioner’s Office. 2015. Available at: https://

ico.org.uk/. Accessed April 30 (2015).14. A. B. Munir and S. H. M. Yasin, Personal Data Protection in Malaysia: Law

and Practice, Sweet and Maxwell Asia, Selangor (2010).15. C. Millard, Cloud Computing Law, Oxford University Press, Oxford (2013).


3351




Vol. 21, 3356–3359, 2015

Service Orientation Initiative Process Towards

Enterprise Services Environment

Muhammad Suhaizan Sulong1�∗, Azlianor Abdul-Aziz1, Andy Koronios2, and Jing Gao2

1Faculty of Information and Communication Technology, Universiti Teknikal Malaysia Melaka, 76100, Malaysia2School of Information Technology and Mathematical Sciences, University of South Australia, Australia

Enterprises are now seeking to cater the needs of their business change to improve business agility and toincrease information technology flexibility. They have transformed with the initiative of implementing serviceorientation from their existing information technology infrastructure to a managed services environment. Thisrequires a step-by-step process in the implementation of the service orientation initiative. In this paper, the four-phase process of implementing service orientation initiative derived from the related literature is taken to thecase study of seven case enterprises to check for its applicability in wider enterprise settings. It has resultedthat this four-phase service orientation initiative process has been clearly employed and agreed upon by mostof the case enterprises. It is therefore believed that the four-phase process can be a generic initiative processof implementing service orientation for any enterprises to maintain viability in a services environment.

Keywords: Service Orientation, Enterprise Services, Service-Oriented Architecture.

1. BACKGROUND STUDYMany enterprises nowadays have deployed and integrated multi-

ple information systems into their core business operations across

different departments.1 These information systems frequently

shared and produced similar information to support business

objectives across the departments. This shows that these informa-

tion systems implemented are impractical and developing more

of this can lead to high costs. Besides, as business conditions

and environment changes with evolving consumer needs, the

enterprises are seeking competitive advantages through advanced

information systems that should reflect the changes.2 The current

information technology (IT) architecture that is siloed mololithic

in style could not support the advanced IT requirements.3 Thus,

rapid changes in the business environment have made planning

for new architecture more important and are required to reduce

unnecessary costs from using the existing information systems.

The new architecture is referred to a service-oriented architecture

i.e., service orientation in which enterprises that are having an

initiative to implement this would increase IT flexibility as well

as improve business agility.3

Service orientation as well as service-oriented architec-

ture (SOA) has been a buzzword for enterprise seeking services

environment. A clear understanding of what service orientation

means is important from two perspectives. From a technical

perspective, it can be referred to as a technology for enabling


business4 which is defined as “architectural style where systems

consist of service users and service providers.”5 While from the

business perspective, it can be referred to as aligning technol-

ogy with business,6 which is defined as a paradigm or a way

of thinking or designing information systems.7 We consider both

definitions for this study.

Implementing service orientation initiative at the enterprise

level needs to be in phases through a step-by-step approach8

which means this enterprise initiative transforms to services envi-

ronment that can quickly respond to business change. A recent

study has reviewed the related literature on the service orientation

initiative extensively.9 From the study, five service orientation

initiatives are discussed and evaluated (refer to Table I). Both

Oracle10 and IBM11 have, in reality, been in the service orienta-

tion arena for a long time and provide own unique approaches

to comprehensive service orientation in which they retrofit SOA

into their established computing platforms. Gartner Research

also extends its specific approach namely Application Activity

Cycle for SOA12 to implementing service orientation initiative.

Although the service orientation marks an active area of devel-

opment in industry, the researchers in academia also intensified

their efforts to realise the potential of service orientation in which

they proposed SOA Roadmapping13 and SOA Adoption Man-

agement Roadmap.14 It is worth noting that in order to increase

confidence of the review result, a greater variety of service ori-

entation initiatives is covered.

After having carefully reviewed and compared these five

service orientation initiatives, they have a high integrity and



consistency regarding the service orientation process and thus,

the general four phases9 are derived:

• Strategic Planning: Starting point of an enterprise to plan

strategically its service orientation initiative.

• Building Architecture: Designing and developing services and

infrastructure to integrate with existing operating platforms.

• Services Operation: All developed services are being exposed

to allow integration with other systems and services.

• Continuous Improvement: Service orientation is at the point

of improvement either changing current services or creating new

services to accommodate business change and measure its per-

formance.

The name of the second phase is suggested to rephrase to

“Architecture Building” for a proper terminology.

2. RESEARCH METHODOLOGYThis study used a multiple case study design15�16 that involve

analysing real-world case situations of seven enterprises with var-

ious types of business in order to seek how these enterprises

implementing SOA initiative to increase their business agility.

From these seven enterprises, 28 participants, in their respec-

tive SOA team members grouped into three levels; management,

architectural and delivery, were gave written informed consent

for the semi-structured interviews to gain in-depth insights into

how they experienced the service orientation initiative process.

These seven enterprises were selected on an opportunistic basis

which means the selection based on any invited enterprises that

having a service orientation initiative and agreed to participate in

this study.

The format of each interview was one to one session lasted

approximately an hour, which sought to create a relationship

between the researcher and the participant, providing dynamism

Table I. Comparison of service orientation initiative process.9

Service orientationStudy initiative process

Oracle’s approach to SOA10 1. Establish a strategic plan for SOAadoption; 2. execute the SOA programlevel activities; 3. deliver projects andservices following SOA best practices;4. establish ongoing guidance andgovernance.

IBM scope of SOA adoption11 1. Ad hoc stage; 2. technology adoptionstage; 3. line-of-business adoptionstage; 4. enterprise adoption stage;5. value-net adoption stage

Gartner application activitycycle for SOA12

1. Strategise: SOA adoptionconsiderations; 2. evaluate: planningand designing SOA systems;3. execute: implementing andmanaging SOA in the real world;4. review: improving and refining theuse of SOA; 5. innovate

SOA roadmapping13 1. Planning and analysis; 2. design andconstruction; 3. deployment andoperations; 4. management andgovernance

SOA adoption managementRoadmap14

1. Set strategy organization; 2. planningstrategy; 3. organization and business;4. operations planning; 5. design;6. implementation; 7. monitoring andtesting; 8. establishment and feedback.

and flexibility in the discussion.17 Table II presents the back-

ground of participating enterprises, including their service orien-

tation initiatives (implementation year, platform and system) and

volunteered participants (job title).

The analysis was assisted with the use of analytical software

tool called NVivo.18 This tool can assist in categorising and

analysing the interview data upon transcription. The analysis pro-

cess includes first is to transcribe the interview data. Next, using a

thematic approach, the interview data can be categorised and then

interpreted towards the research objective. The findings across

the cases regarding the process of their service orientation initia-

tive implementation are presented in the next section.

3. RESEARCH FINDINGS DISCUSSIONThe general four-phase of the service orientation initiative

process—strategic planning, architecture building, services devel-

opment and continuous improvement—has been taken as a

basis for conducting case study. This general service orientation

Table II. Case enterprises background.

Background

Case SOA program Participant

A Initiated 2009 on IBMplatform; consumersystems for integrationacross the enterpriseinclude information, datamodelling to developnecessary services.

Head of SOA; Head ofArchitecture and Strategy;Lead Architects (3);Solution Architects (2);Service Modelling andDesign; EngagementManager.

B Initiated 2010 on IBMplatform; deliver ITsolutions and infrastructureglobally using SOAapproaches to improvearound the level of reuse.

Head of Architecture; SolutionArchitect; Head ofIntegration Services;Delivery ManagerIntegration.

C Initiated 2005 on IBM platform;Centralised control of itspayroll system combinedwith distributed accessthroughout the country.

Solution Architect; Team LeadSolution Design; ManagerApplication Development.

D Initiated 2007 on Oracleplatform; Implement SOA intheir customer, retail andfinancial systems.

Project Manager; SolutionArchitect; IntegrationProgram Manager

E Initiated 2010 on Microsoftplatform; A federally fundedproject for implementingSOA specific in GISsystems for naturalresource data.

Chief Information Officer;Operations Manager;GIS Administrator

F Initiated 2008 on software AGand IBM platform; To linktogether disparatesystems—To support theacquisition strategy forfuture integrationrequirements.

Integration Design Manager;Senior DevelopmentManager; DevelopmentManager

G Initiated 2004 on SAPplatform; A plan to developsophisticated enterprisesoftware applications inretail systems to connectto CRM.

Enterprise Architects (2);Solution Architect

Note: ( )—denotes number of participants.

3357


Table III. Summary of service orientation initiatives implementation across cases.

Similar with general approach Different approach

Case A Case C Case D Case F Case G Case B Case E

Clearly stated thatthey have plan,build, operateand continuousimprovement.

Clearly stated thephases—Planning theintegration, builtsome services,deployed andcontinuouslyimproved them.

Clearly stated thatthey wentthrough thesefour phases togeneralise allthe services ofall application.

Clearly stated thatthey plan anddetermineservices toexecute on.Then create theservices,implement andmeasure them.

Clearly statedthat—The fourphases tookplace toimplement all ofthe top-downservices thatwere required.

Use the IBM’smethod—Service-OrientedModelling andArchitecture(SOMA).

Via an outsourcedcompany usingthe Microsofttechnology.

initiative was agreed upon by most cases. Although the findings

can be depends on the implementation platform, most respon-

dents stated that their service orientation initiatives have imple-

mented through this general approach.

That makes a great deal of sense. We fell into a holebetween planning and building. We have plan, buildand operate it all happening at once and continu-ous improvement too. � � �There are still many otherthings that we are, I guess planning and building,and of course the operational team is still working.So there’s actually the operations guys are contin-ually improving their capabilities as well.

Enterprise Infrastructure Architect (Case A)

The phases, yeah, we have done some planning ina very limited area of the integration. We have builtsome services on the hub so that they can be usedand we have actually deployed them and we areusing them. Have we continuously improved them?Not to my knowledge.

Team Lead Solution Design (Case C)

Well, we went through these different phases andwe experiment the four phases, because we couldonly trying to extend the model to generalise all theservices of all applications.

Integration Program Manager (Case D)

We’d do that. So you will do your plan. Youwill then determine, understand what services, etc.you’re going to execute on. You will then create theservice, implement, measure your ROI.

Senior Development Manager (Case F)

Yeah, it pretty much happened like that. Like theprobably the only thing for a particular system ora particular service, [we] tried to implement all ofthe top-down services that were required.

Application Architect (Case G).

Only two cases—Case B and E, indicated they had used SOA

vendor’s methodology to implement their service orientation ini-

tiative. For the case B enterprise, as they used IBM for their

service orientation initiative implementation, the methodology

that IBM provides was SOMA—Service-Oriented Modelling and

Architecture. According to the respondent:

We’re definitely looking to continue to implementSOA in terms of our progress along the SOA model,

we have done a lot of things to try and mature theway that we use SOA. So we use the SOA method-ology, SOMA. � � � If you follow �our� lifecycle for aproject you get an initiation, where people are try-ing to think of a technology solution. They then gointo a defined phase, where they’re trying to findwhat that solution is, and then they go into an anal-ysis and design phase. So we’re a delivery team—allwe do is to deliver.

Head of Integration Services (Case B)

Besides, in the Case E enterprise, their service orientation

initiative has implemented using Microsoft technology with the

assistance of an outsourced company as the respondents stated:

This organisation was ripe for an SOA, for animplementation of an SOA to meet some departmen-tal objectives and government objectives and so thatwas really what initiated that. It was actually out-sourced to an Indian company. So with regards tothe actual SOA technology and infrastructure that’sbeen built on, I was making sure when they weredeveloping and designing its functionality, it metour needs.

GIS Administrator (Case E)

It’s trying to be Microsoft based at this stage, whichis Microsoft Dynamics.

Operations Manager (Case E)

Although the enterprises of Case B and E followed IBM and

Microsoft’s approaches in implementing their service orientation

initiative respectively, the approaches also have the generic pro-

cesses like capturing user requirements during the product design.

Table III shows the summary of cases that clearly stated in

following the general approach and cases that have different

approaches in implementing their service orientation initiatives.

Therefore, it is clear that the proposed general approach via the

four-phase implementation can be employed for implementing

service orientation initiatives. One important implication is that

any enterprises who are still pursuing their service orientation

initiatives can consider the use of this general approach through-

out its implementation process in order to become SOA-enabled

enterprises.

4. CONCLUSIONService orientation is significant to enterprises in order to

improve business agility and IT flexibility as well as achieve the

3358


business value of SOA. The service orientation initiative pro-

cess is developed generally by comparing five different service

orientation initiatives studied previously. The general phases are

strategic planning, architecture building, services operation and

continuous improvement. From the Table III, five out of seven

case enterprises have affirmed that they applied the general ser-

vice orientation initiative process when implementing their ser-

vice orientation initiatives. Detail explanation for each phase

according to the case study is as follows:

• Strategic Planning: This first phase is where an enterprise

starts to plan its business transformation initiative to service ori-

entation. It involves top-management level that is significant for

supporting the entire initiative implementation in order to achieve

specific business needs. With an implementation strategy that fits

well within the enterprise settings, the service orientation can be

achieved.

• Architecture Building: This second phase is the key to deliver

service orientation into all enterprise operations and systems, in

order to be more agile and efficient, through its strategic partner-

ship of implementation platform. This involves a dedicated team

designed for SOA that can collaborate between all stakehold-

ers to design, develop and govern the entire service orientation

initiative implementation that includes service components and

infrastructure.

• Services Operation: This third phase is the core of the ser-

vice orientation initiative: the operating of loosely coupling new,

reusable and existing software systems and services in a single

integration system of the legacy architecture. By referring to the

operational guideline, enterprise can ensure the service orienta-

tion works accordingly and quickly reacts to necessary business

change inline with the change management plan.

• Continuous Improvement: This fourth phase is at the point

where service orientation continuously improves the full range

of services in order to satisfy the requirements of core business

processes. It also includes in which the performance of service

orientation is measured and updated across the enterprise for

effectiveness.

From this research, we conclude that the general four-phase

service orientation initiative process derived from the literature

and has been affirmed by the case study is a suitable process to

implement the service orientation initiative in leading to becom-

ing a full services environmental enterprise.

Acknowledgments: This research is funded by the Uni-

versiti Teknikal Malaysia Melaka (UTeM) and we gratefully

acknowledge its generous financial support.

References and Notes1. M. P. MacDonald and D. Aron, Leading in Times of Transition: The 2010 CIO

Agenda, Gartner Inc., Report (2010).2. Y. Gong and M. Janssen, Government Information Quarterly (2011).3. J. Choi, D. L. Nazareth, and H. K. Jain, Journal of Management Information

Systems 26, 253 (2010).4. S. Bistarelli and P. Campli, Fairness as a QoS measure for web ser-

vices, Young Researchers Workshop on Service-Oriented Computing (2009),pp. 115–127.

5. P. Bianco, R. Kotermanski, and P. Merson, Evaluating a serviceoriented archi-tecture, Software Engineering Institute, Technical Report (2007).

6. H. M. Chen, R. Kazman, and O. Perry, IEEE Transactions on ServicesComputing 3, 145 (2010).

7. T. Kokko, J. Antikainen, and T. Systä, Adopting SOA—Experiences from nineFinnish organizations, European Conference on Software Maintenance andReengineering (2009), pp. 129–138.

8. G. Feuerlicht and S. Govardhan, SOA: Trends and directions, Proceedingsof 17th International Conference on Systems Integration, Czech Rep. (2009),pp. 149–155.

9. M. S. Sulong, A. Koronios, J. Gao, and A. Abdul-Aziz, Implementing qual-ity service-oriented architecture initiative in organisations, Portland Interna-tional Conference for Management of Engineering and Technology (2012),pp. 3666–3673.

10. B. Hensle, Oracle’s approach to SOA, Oracle Corporation, IT Strategies fromOracle: Data Sheet (2010).

11. A. Arsanjani and K. Holley, Increase flexibility with the service integra-tion maturity model (SIMM): Maturity, adoption, and transformation to SOA.IBM.com (2005).

12. R. Altman, SOA overview and guide to SOA research. Gartner Inc., Report(2010).

13. T. C. Shan and W. W. Hua, Service-oriented architecture roadmapping,Congress on Services-I, CA, USA (2009), pp. 475–476.

14. A. Moeini, N. Modiri, and T. Azadi, Service oriented architecture adoptionmanagement roadmap, 7th International Conference on Digital Content, Mul-timedia Technology and its Applications, Iran (2011), pp. 119–124.

15. K. M. Eisenhardt, Academy of Management Review 14, 532 (1989).16. K. M. Eisenhardt and M. E. Graebner, Academy of Management Journal

50, 25 (2007).17. A. Fontana and J. H. Frey, The interview: From structured questions to negoti-

ated text, Handbook of Qualitative Research, 2nd edn., edited by N. K. Denzinand Y. S. Lincoln, Sage Publications, CA (2000), p. 645.

18. A. Lewins and C. Silver, Using Software in Qualitative Research: A Step-by-Step Guide, Sage Publication, USA (2007).


3359




Vol. 21, 3373–3377, 2015

Teachers Acceptance of Frog Virtual

Learning Environment (E-Learning):

Case Study of Vocational College

Nurbaya Mohd Rosli1, Nurazean Maarop2�∗, and Ganthan Narayana Samy2

1Kolej Vokasional ERT Setapak, Kuala Lumpur, Malaysia2Advanced Informatics School, Universiti Teknologi Malaysia, Jalan Yahya Petra, 54100 Kuala Lumpur, Malaysia

E-learning has been implemented by education institution worldwide for decades, including Kolej VokasionalERT Setapak, under the management of Ministry of Education, Malaysia. Implementing a successful e-learningnamely FROG VLE, is another step of collaborating education with technology, thus, it is important for collegeto recognize the acceptance factor of FROG VLE among teachers. This study aims to explore the acceptancefactors namely, perceived usefulness, perceived ease of use, computer self-efficacy, convenience, champion’scharacteristics, instructional design, and technological factor towards behavioral intention as technology accep-tance indicator. The research is based on mix-method study involving semi-structured interviews and question-naires survey. All these factors were explored to examine factors that influence the acceptance. The resultsindicated that perceived usefulness, perceived ease of use, instructional design, convenience, technologicalfactor and computer self-efficacy have significant effect to the acceptance of FROG VLE. However, champion’scharacteristic was found quantitatively insignificant towards the acceptance of FROG VLE but bear relevantto be considered as important from qualitative standpoint. Finally, providing e-learning technology acceptancefactors may help education institution to implement and enhance the technology for efficient use of e-learning.

Keywords: Technology Acceptance, E-Learning, School Online Learning.

1. INTRODUCTIONDevelopment in Information and Communication Technologies

(ICT) has its impact on education sector in Malaysia. E-learning

is an innovative approach for delivering electronically mediated,

well-designed, student-directed and interactive learning environ-

ment for everyone, regardless of time and place, using either the

Internet or digital technologies in collaboration by the principles

of instructional design.1 In Malaysia, Learning Management Sys-

tem (LMS), an e-learning, has been implemented by pilot project

since 2009 using the open source technology, LMS Moodle.2

E-learning is defined as learning facilitated and supported

through the utilization of information and communication tech-

nologies (ICTs).3 Effective implementation of an e-learning

initiative requires attention to a number of issues including tech-

nological, pedagogical and individual factors.4 Besides, good

connectivity is very important in order to realize the functionality

of e-learning used by the community.

In 2010, through Bahagian Teknologi Pendidikan (BTP),

Ministry of Education (MOE) has implemented an e-learning;

“Learning Management System” (LMS) to 50 pilot schools

throughout Malaysia. According to Assistant Director, Learning


Resources Sector of BTP, the constraints when implementing

LMS are the internet connection problem and users were com-

plaining on the interface of the e-learning. There are eight out

of 50 pilot schools do not use LMS in implementation within

10 months, likely the LMS system in the school is not con-

nected with BTP server.2 In 2012, a new e-learning system

known as FROG VLE was introduced. The system is a cloud-

based learning platform that can be accessed by teachers, stu-

dents and parents. This project is currently been implement in all

school in the country by using the 4G internet technology under

1BestariNet program.

In developing countries, technological factors still remain as

an obstacle in implementing online learning system, where the

advancement of IT infrastructures development countries is far

behind from their developed counterparts.5�6 Therefore, this study

explores the acceptance using appropriate acceptance model.

In regard to the context of this study, 1BestariNet is an “End

To End” (E2E) network service for the purpose of teaching

and learning process and for management and administration

of all 10,000 schools under Ministry of Education, Malaysia.

This project is the main platform for the proposed e-learning

concept (FROG VLE). The implementation of this project is to

enhance the internet access bandwidth for all school and the



main component in 1BestariNet project is the Virtual Learning

Environment (VLE).

The significance of this study is to explore teachers’ accep-

tance on FROG in the context of vocational school in Malaysia

by applying an appropriate model to help explain the acceptance

scenario. This study would reveal those factors that influence

the acceptance of FROG VLE among teachers at Kolej Voka-

sional ERT Setapak (KVertS) and how e-learning help teachers

in the teaching and learning process. Furthermore, the result of

this study can be benchmarks for the college in considering a

solution for the institutions. Besides, this study can also enhance

and contribute to the technology acceptance body of knowledge.

2. STUDY BACKGROUNDVirtual learning environment (VLE) is a web-based communica-

tions platform that allows students, without limitation of time and

place to access different learning tools such as program informa-

tion, course content, teacher assistance, discussion boards, docu-

ment sharing systems and learning resources.7�8 VLEs are rapidly

becoming an integral part of the teaching and learning process.8�9

Thus, it enables improvements in communication efficiency, both

between student and teacher, as well as among students.7�8

User acceptance has been viewed as the pivotal factor in deter-

mining the success or failure of any information system project.10

Most of this work been focused on productivity-oriented or util-

itarian systems.11 Therefore, user acceptance is defined as will-

ingness to employ IT for the tasks it is designed to support.12

Throughout the study, there are several models have been devel-

oped to investigate and understand the factors involving the

acceptance of computer technology in organizations. The models

are about to study the user acceptance, adoption and IT usage

behaviour.

Virtual Learning Environment (VLE) is known as a good plat-

form of e-learning, specifically in this study for FROG. The study

from various journals show many researchers have been review-

ing on the acceptance and adoption of e-learning using different

types of technology acceptance model. The studies used the com-

bination of keywords in searching the acceptance theories and

related e-learning acceptance. Most of search is from Science-

Direct and IEEE databases.

From the literature review of e-learning acceptance, it

shows that most researchers are using Technology Acceptance

Model (TAM)13 as their research model. According to William

and Jun,14 TAM is a powerful and robust predictive model which

results from the meta-analysis of technology acceptance model

study.

One of the researches in the e-learning acceptance area was

done by Hussein et al.5 as shown in Figure 1 and they used TAM

as their base theoretical model. The objective of their study was

to investigate the factors that affect acceptance of e-learning. This

research was conducted using quantitative methodology where

the respondents were online students of the Indonesian Open

University (UT) from various major of study. As a result, this

research had support instructional design, computer self-efficacy

and technological factor as predictors of e-learning acceptance in

a developing country. According to Šumak,15 TAM has been used

significantly to explain technology acceptance in schools either

among students or teachers. Hence, this research has considered

appropriate model to help explain the acceptance of FROG VLE

in the context of the study.

InstructionalDesign


TechnologicalFactor

Convenience

Instructor’sCharacteristic

PerceivedUsefulness

PerceivedEase of USe

Intention toUse

Fig. 1. Original model by Hussein et al.5

In regard to the current environment of FROG VLE implemen-

tation, there was no presence of instructor in the environment.

However, e-learning champion had been selected to assist and

share their skill and knowledge. Hence, instructor characteristic

from the original model as shown in Figure 1 has been replaced

with champion’s characteristics as an external factor in this study

context.

Self-efficacy is defined as the judgment of one’s capability to

use an IT or computer.17�18 Previous studies have shown that

computer self-efficacy is related to technology acceptance.5�18

The design of online learning is similar type of classroom

format where there are course description, objectives, content,

purpose, scope and evaluation.19 According to Hussein et al.5

a well-designed application is believed to have an effect on online

learning adoption. Therefore, teachers must be able to easily find,

read, download and save the materials that been shared online.

Technological factors still remain as an obstacle in imple-

menting online learning system in developing countries, where

the advancement of IT infrastructures is far behind from their

developed counterparts.20 According to Peters (2002),21 problems

with connection, low modem speed, missing links, loading page

and availability of memory are some of the obstacles in online

learning.

Convenience is one of the enabling factors identified in the

online learning literature [5]. According to Tobin (1998),22 con-

venience is achieved when students can access the learning at

convenience time. For this study, teachers can implement a class

whenever and wherever they want, as the students will access

accordingly.

Champion’s characteristic is another factor believed to influ-

ence teacher’s acceptance of FROG VLE. A successful imple-

mentation of e-learning does not only rely on advanced

technology, but also champions as a reference key people, who

is responsible to share their knowledge and encourage teachers

to use FROG VLE. According to a study in Australia in 2009,22

champion can be regarded as an agent that can set the learning

of an e-learning where champions is skilled in e-learning, willing

to share their expertise with passion and enthusiasm and willing

to solve problems, either technical or non-technical problems.

Their study discovered that e-learning champions have the abil-

ity to empower, motivate and mentor teachers in e-learning and

establish effective networking with other teachers to encourage

knowledge transfer and technology exploration.22 Figure 1 illus-

trates the altered research model by Hussein et al.5 considering

3374


champion characteristics whereas the overall proposed model to

be used in this study was shown in Figure 2.

In summary, the theoretical background of this study is

based on Technology Acceptance Model with external factors by

Hussein et al.,5 where the research evidently support the instruc-

tional design, technological factors and computer self-efficacy

factors as a very important role in acceptance towards e-learning.

However, convenience and instructor’s characteristics were not

dominant factors in the research, therefore, for this study, cham-

pion’s characteristic was replacing instructor’s characteristics.

Thus this is in line with the recommendation from the previous

study to look into other factors which were not addressed by

their.

3. METHODOLOGYThe study used mixed-method research design.23 In particular,

in gathering the data, a mixed-method convergent parallel design

introduced by Creswell and Clerk (2011)23 was implemented to

compare or relate the data and followed by interpretation from

the result gained. Questionnaires and interview questions are pre-

pared to comply with the studies. The purpose of this approach

was that quantitative data and subsequent appropriate analyses

would provide a strong understanding to the whole results.23�24

Five teachers participated in interview whereby thirty teachers

representing the 83% population of FROG VLE participated

in the questionnaire survey. In designing the questionnaire, the

study used the five likert scale questions to obtain the level of

agreement or disagreement. Seven out of eight factors used in

the survey for this study were adapted from the previous study

and one factor is included for the context of this study.

4. RESULTIn regard to quantitative analysis, descriptive analysis and the

Spearman correlation were applied for quantitative analysis.

On the other hand, the qualitative findings of this study were

based on thematic analysis and the presentation of findings

were based on weighting-the-evidence approach by Miles and

Huberman.25 The results of the study are as follows.

4.1. Quantitative Result

Table I shows the correlation coefficient analysis result for the

study. From the table, it illustrates the association between BI

and all variables except CC were considerably strong (rs > 0�50),

while CC was not significant as the p-value is more than 0.05.

PerceivedUsefulness


PerceivedEase of Use

Intentionto Use

InstructionalDesign

TechnologicalFactor

Convenience

Instructors’Characteristic

Fig. 2. Proposed model.

Table I. Correlations of factors.

Correlation strengthBI and significance

BICorrelation 1.000 –Sig. (2 tailed) 0.000

PUCorrelation 0.794 Strong and significantSig. (2 tailed) 0.000

PEOUCorrelation 0.751 Strong and significantSig. (2 tailed) 0.000

CSECorrelation 0.507 Strong and significantSig. (2 tailed) 0.004

CCCorrelation 0.417 Not significantSig. (2 tailed) 0.022

CONVCorrelation 0.801 Strong and significantSig. (2 tailed) 0.000

TFCorrelation 0.674 Strong and significantSig. (2 tailed) 0.000

IDCorrelation 0.784 Strong and significantSig. (2 tailed) 0.000

4.2. Qualitative Result

The findings of the study of perceived usefulness indicated that

this factor shows that FROG VLE can improves and enhance

teaching and learning activities. Majority of the respondent stated

about the positive impact when they used FROG VLE during

their teaching process. Example of excerpts:

“� � �FROG VLE helps in my teaching performance,where I can use notes that been share with otherteachers and I can surf any new materials throughfrog sites and frog store � � �” �Participant 1�

“� � � it improves my quality of teaching, from tra-ditional style to the use of ICT especially for myteaching materials � � �” �Participant 5�

In this study for variable perceived ease of use, all respon-

dents represent that perceive ease of use supports the use of

FROG VLE. Respondent 5 finds FROG VLE is quite easy, but

need more time to understand the e-learning functions, especially

the widgets and emphasis that the lack of exposure and using

online applications affects the understanding of using FROG

VLE. Example of excerpts:

“� � � It is quite easy for me, but I need sometimesto understand the functions especially the use ofwidget � � �” �Participant 3�

“� � �The widget ease me, just drag from the menu.Plus, I use Google drive to import material intoFROG � � �” �Participant 1�

“� � � It is easy. More or less is similar to FBfunctions � � �” �Participant 2�

The majority of the respondents show that they are IT savvy

user. Most of them are exposed to use computer in their work,

specifically in teaching. However, based on counting measure-

ment, From the descriptive analysis, the majority of the users

3375


have ticked for “Neutral” in computer self-efficacy, this conclude

that, teachers in KVertS need guidance to use FROG VLE. Exam-

ple of excerpts:

“� � �No. At the first time, I need someone who skill-ful to use it to assist me � � �” �Participant 3�

“� � �Generally, I used to download and save filefrom online site, so I don’t have problem on it.I’m confident � � �” �Participant 4�

“� � �No. I seldom use application online, soI need someone to assist me in using FROG � � �”�Participant 5�

During the interview, interviewee were asked about whether

FROG champion’s in the college really help in exploring and

use the e-learning, the results were exactly the same. The FROG

champions are skillful and willing to share their expertise,

besides they are friendly and approachable. Example of excerpts:

“� � � the champion is very good. He is very help-ful and eager to guide me to use FROG � � �”

�Participant 5�

“� � �They are eager to share whatever they know.If I asked them via phone, they still giveguidance � � �”

In regard to technological factor, all respondents emphasized

that, poor internet connections demotivate them to use FROG

VLE in their teaching activity.

“� � �The interface is good and user friendly. If theinternet connection is poor, it gives problem � � �”�Participant 1�

“� � � Interface is fantastic and user friendly. Withoutgood internet connection, it takes so long for thepage to load. This is frustrating � � �” (Participant 2)

4.3. Overall Mixed-Method Finding

The results of both interviews and questionnaires analysis are

shown in Table II.

In regard to Perceived Usefulness, the results of both data are

consistent. Teachers agreed that (PU) affects their acceptance of

FROG VLE.

In regard to Perceived Ease of Use, The results of both data are

consistent. Teachers agreed that (PEOU) affects their acceptance

of FROG VLE.

In regard to Computer Self-Efficacy, the results of both data are

consistent. Teachers agreed that (CSE) affects their acceptance

of FROG VLE.

In regard to Champion’s Characteristics, the results of both

data are not consistent. Spearman’s correlation result is not sig-

nificant for (CC).

In regard to Convenience, the results of both data are consis-

tent. Teachers agreed that (CONV) affects teachers’ acceptance

of FROG VLE.

In regard to Technological Factors, the results of both data are

consistent. Teachers agree that (TF) affects teachers’ acceptance

of FROG VLE. Even though the current state of technology provi-

sion is not favorable but the respondents agreed TF is very impor-

tant for usage continuity. Further, the correlation indicates strong

relationship between TF and the acceptance of FROG VLE.

Table II. Concluding result based on mixed-method.

Interview Questionnaire ConcludingFactors responses results result

Perceivedusefulness

StronglyRelevant

Significant correlationwith 0.794

Significant.

Perceived easeof use

Relevant Significant correlationwith 0.751

Significant.

Computerself-efficacy

Relevant Significant correlationwith 0.507

Significant.

Champions’characteris-tics

StronglyRelevant

Not significantcorrelation with0.417; sig. (2-tailed)at 0.022

Less significant.

Convenience StronglyRelevant


Significant.

Technologicalfactors

StronglyRelevant


Significant.

Instructionaldesign

StronglyRelevant


Significant.

Behavioralintention

StronglyRelevant

Significant value forall factors thatassociate with (BI)except (CC)

Significant.

In regard to Instructional Design, the results of both data are

consistent. Teachers agree that (ID) affects teachers’ acceptance

of FROG VLE.

Overall, the FROG VLE is accepted by teachers in KVertS as

indicated by Behavioral Intention.

5. CONCLUSION AND DISCUSSIONThis study has identified factors influencing the acceptance of

education technology in the vocational school environment. The

study has further explored the acceptance of FROG VLE using

the proposed model. In order to obtain more beneficial results in

future research, this study suggests the following aspects.

• Further researches may extend the scope of study by including

samples from different educational institution such as primary

school or secondary school.

• Include more elements and factors that affect the acceptance

of FROG VLE among teachers in the research to achieve more

understanding regarding FROG VLE acceptance.

• Use other acceptance models to explore the acceptance of

FROG VLE among teachers.

• Apply observation method for data collection to gain more

results for the research.

Findings obtained from this research shows that perceive ease

of use, computer self-efficacy, convenience, instructional design,

champion’s characteristics and technological factors have sig-

nificant effect towards the acceptance of FROG VLE among

teachers. From this result, educational institution may have more

insight in accepting and implementing e-learning in college or

school for more effective teaching and learning activity.

References and Notes1. J. L. Morrison and B. H. Khan, The global e-learning framework: An inter-

view with Badrul Khan, The Technology Source, A Publication of the MichiganVirtual University (2003).

2. T. S. Wai, Jurnal Pembestarian Sekolah 2010 BTP KPM (2010).3. M. Jenkins and J. Hanson, E-Learning Series: A guide for Senior Managers,

Learning and Teaching Support Network (LTSN) Generic Centre, UnitedKingdom (2003).

3376


4. M. Masrom and R. Hussein, User Acceptance of Information Technol-ogy: Understanding Theories and Models, Venton Publishing, Kuala Lumpur(2008).

5. R. Hussein, U. Aditiawarman, and N. Mohamed, E-learning acceptance ina developing country: A case of the Indonesian Open University, GermanE-Science Conference (2007).

6. W. C. Poon, L. T. Low, and G. F. Yong, The International Journal of EducationalManagement 18, 374 (2004).

7. L. L. Martins and F. W. Kellermanns, Academy of Management Learning andEducation 3, 7 (2004).

8. J. J. L. Schepers and M. G. M. Wetzels, Technology acceptance: A meta-analytical view on subjective norm, Proceedings of the 35th European Mar-keting Academy Conference, Athens, Greece (2006).

9. K. A. Pituch and Y. K. Lee, Computers and Education 47, 222 (2006).10. F. D. Davis, A Technology Acceptance Model for Empirically Testing New End-

User Information Systems: Theory and Results, Sloan School of Manage-ment, Massachusetts Institute of Technology: Doctoral Dissertation (1986).

11. V. Venkatesh and F. D. Davis, Management Science 46, 186 (2000); F. D.Davis, MIS Quarterly 13, 319 (1989).

12. A. Dillon and M. Morris, User acceptance of new information technology:Theories and models, Annual Review of Information Science and Technol-ogy, edited by M. Williams, Information Today, Medford, NJ (1996), Vol. 31,pp. 3–32.

13. F. D. Davis, MIS Quarterly 13, 319 (1989).14. R. William and H. Jun, Information and Management 43, 740 (2006).

15. B. Šumak, M. Hericko, M. Pušnik, and G. Polancic, Informatica 35, 91(2011).

16. R. Agarwal and E. Karahanna, MIS Quarterly 24, 665 (2000).17. M. E. Gist, Personnel Psychology 42, 787 (1989).18. E. E. Grandon, K. Alshare, and O. Kwun, Journal of Computing Science in

College 20, 46 (2006).19. W. C. Poon, L. T. Low, and G. F. Yong, The International Journal of Educational

Management 18, 374 (2004).20. O. Peters, Distance Education in Transition: New Trends and Challenges,

Bibliotheks-und Informationssystem der Universität Oldenburg, Oldenburg(2002), pp. 37–45.

21. K. Tobin, Qualitative perceptions of learning environment on the world wideweb, International Handbook of Science Education, edited by B. J. Fraserand K. G. Tobins, Kluwer Academic Publishers, United Kingdon (1998),pp. 139–162.

22. M. Jolly, B. Shaw, K. Bowman, and C. McCulloch, Final report—The impactof e-learning champions on embedding e-learning—In organisations, indus-try or communities, Department of Education, Employment and WorkplaceRelations, Australian Government (2009).

23. J. W. Creswell and V. L. P. Clark, Designing and Conducting Mixed MethodsResearch, Sage, Thousand Oaks, CA (2011).

24. N. Maarop and K. T. Win, Journal of Medical Systems (J. Med. Syst.) 36, 2881(2012).

25. M. B. Miles, A. M. Huberman, and J. Saldana, Qualitative Data Analysis:A Methods Source Book, Sage Publication (2013).


3377




Vol. 21, 3382–3384, 2015

A Review on the Developing Algebraic Thinking

Najihahbinti Mustaffa∗, Zalehabinti Ismail, Zaidatunbinti Tasir, and Mohd Nihra Haruzuan Bin Mohamad Said

Faculty of Education, University Technology Malaysia, 81310, Skudai, Johor, Malaysia

Algebra is interrelated with other mathematical topics such as statistics and geometry. Learning algebra is aboutsolving an unknown and is the way of thinking. Algebraic thinking is a process of thinking in specific domain inMathematics. Teaching and learning process of algebra should be reform nowadays as to beyond computationaland assisted with technology. This paper is presented that algebraic thinking should be developed throughtechnology concurrently with the role of teacher. The study of developing algebraic thinking through technologyintegrated with the role of teacher should be conducted.

Keywords: Algebraic Thinking, Technology, Teacher.

1. INTRODUCTIONAlgebraic thinking is the beyond of algebra which consists of

variables and in terms of algebraic expressions. Algebraic think-

ing is the way of solving x and y, which makes students think

abstractly and make reasoning in real-life problems. Components

of algebraic thinking consists of reasoning of unknown, able to

generalize and understand the concept of equal sign.2 Algebraic

thinking is a process of thinking algebraically. It requires students

to think and solve problems using abstractions logically.3 Alge-

braic thinking is about the ability of student to analyze, make

generalization, able to solve the problem, predict, justify, prove

as well as able to notice the changes and modeling.11

Algebraic thinking should start at the early grade and it is also

a process of thinking in a specific domain.

2. PROBLEMS IN DEVELOPING

ALGEBRAIC THINKINGStudents lack of basic skills in algebra commit errors in solving

algebraic problems, weak in the simplification of equations and

algebraic expressions, as well as the interpretation of quadratic

graphs.1 These students see the irrelevant of arithmetic to alge-

bra and are unable to connect these two aspects in theirlearn-

ing process.4 Therefore, the students just memorize what they

have learnt in algebra with thinking. They are only able to solve

lower-level thinking problems because they memorize facts and

algorithms.15 Algebra as the gateway to the high school and

university level is not resolved until nowdue to the readiness

of students to learn through extensive practice with algorithms.

Worksheets, practice exercises and factual questions will not sup-

port the deepening of students’ thinking and only focus on the

memorization and computational procedures.6


Students also face problems when they solve the operations as

objects, variables, functions, and invariable relations or structures,

as well as procedures.11 Various methods were applied to solve

these problems including the usage of technology, particularly

because a pencil-and-paper learning environment does not lead to

the linking between symbolic and algebraic manipulations.15

However, the usage of technology is time consuming. Students

faced the limited access to practice and lack of assessment

through technology hinders the students from acquiring by-hand

skills.14 Furthermore, there is a lack of usage of computers in

mathematics classrooms or labs, despite the students having the

opportunity to choose when and how to use them.15

3. IMPORTANCE OF ALGEBRAIC THINKINGAlgebraic thinking has been emphasized in middle school will

lead to beyond computational aspects and applied in other math-

ematical domains. Algebraic thinking will allow students to think

beyond computational algebra and can connect to the knowl-

edge of future algebra.3 Algebraic concepts should start in the

early age, which require the connections between arithmetic and

algebra with concrete methods.15 Algebraic competence is impor-

tant to be recognized during the process of problem and able

to be solved through algebraic approach, verbally, graphically

or numerically.19 The following are the uses of technology in

previous studies in algebraic thinking classroom, summarized in

Table I.

4. THE DEVELOPMENT OF

ALGEBRAIC THINKINGBased on previous studies,7�18 most of the development of alge-

braic thinking used problem-solving approach. However, sev-

eral elements should be emphasized in problem-solving, such as



Table I. List of published researches on technology.

Author Software Result Level

(Baltaci and Yildiz, 2015) Geogebra 5.0 version Program is usable. Facilitate learning aboutthree-dimensional objects. Can help the acquisitionof age-dependant cognitive level of learners tounderstand three-dimensional objects at an earlierage. Use their time effectively using Geogebra

Pre-service elementaryteacher

(Tabach et al., 2013) Spreadsheet Understanding of the situation and ease of access tothis generational activity. Shows flexible use ofexpressions in a spreadsheet environment from thevery beginning of the school year. Able to generateonly multi-variable and recursive expressions. CIEapproach to beginning algebra can facilitate thetransition from arithmetic to algebra

Primary school students

(Hall and Chamblee, 2013) Geogebra User friendly. Understand the concept of geometricconcept, transformation, and rotation. Enable tosee the changes of equation when the parabolamoves around. Show the changes result onequations

Pre-service teacher

(Bokhove and Drijvers, 2012) MLwiN 2.22 with estimationprocedure RIGLS

Improvement in score Improvement to recognizepatterns Have a sense for symbols

Middle school students

(Bills, Wilson, and Ainley, 2010) Spreadsheet Able to construct meaning for variable. Able toformalize the expression of functional relationships.No kind of thinking that pupils engage during ateaching programme

Primary school students

(Zeller and Barzel, 2010) CAS Influence the student that algebra related toarithmetic Influences the student to evaluatealgebraic representation

Middle school students

(Pierce et al., 2009) CAS CAS worth to be used by most of the students Highperformance students able to manipulate algebraicexpressions Low performance students faceddifficulties to understand the algebraic symbolsand structure of CAS syntax

Middle secondary schoolstudents

the role of a teacher. The significance of the role of a teacher

that is able to ask questions in enhancing students’ thinking is

important. Reversibility, flexibility and generalization of ques-

tions able to develop the generalization as well as deepen stu-

dents’ thinking.6 These types of questions are important as the

questions will lead students to answer, create problems, develop

multiple ways of solution methods, and able to predict the

answers or check the responses from the student. In the learn-

ing environment, students will learn by thinking and not just

memorizing. Students developed their algebraic thinking by gen-

eralizing and relating the problems through sophisticated ways

of thinking such as organizing and manipulation.17 Teachers not

only play a significant role in enhancing students’ thinking, but

technology is also pivotal in this focus.

Nowadays, technology has been used extensively and

widely in learning environment. Algebraic thinking may be

developed using computer environments. Based on previous

studies,3�4�6�8�10�14�15�19 various software have been used to

develop algebraic thinking. Students should be able to understand

the symbols and operations of technology, algebra symbols,4 and

most importantly, the problem of transition between arithmetic

and algebra. CAS can solve the problem of transition between

arithmetic and algebra,19 although it is costly and time consum-

ing. Spreadsheet is applicable only for the basic of generalization.

Furthermore, a previous study5 mentions that for the primary

school students, the thinking engaged in the process of learning

is non-existent. Technology serves as a scaffold in a learning pro-

cess thusit should be efficient, effective, and able to save time in

a learning process such as in learning Geogebra. It must also be

able to improve the view of three-dimensional objects in the early

ages, although this should be tested on specific characteristic of

algebraic thinking.

Computers will help students to connect formal representa-

tions with dynamic visual representations, and allow students

to explore, express and formalize informal ideas. However, the

importance of computers in the learning environment is similar

to the teachers’ role in the classroom. Algebraic thinking requires

problem-solving supported with types of questioning and tech-

nology concurrently.

5. DISCUSSIONThe authors expect that in learning activities of technology, alge-

braic thinking can affect student engagement. In any event, its

intended effect may cause positive or negative impacts. The use

of technology has the potential to be implemented in developing

algebraic thinking, but it will invite challenges and constraints

as mentioned. Viewed from several angles, the authors propose

some alternatives to overcome the constraints and challenges.

Nevertheless, the study on the effectiveness of algebraic think-

ing through technology is still ongoing, and that it may yield dif-

ferent results from those concluded in previous studies. Through

this research, the authors will unlock any issues and constraints

that arise.

6. FUTURE STUDIESThe potential study of algebraic thinking through computers

should be explored. Algebraic thinking software should be

3383


extended to explore the relationships, manipulation of symbols

and procedures, reasoning with representations, using algebra as

a tool and connecting representation. Most of the findings empha-

sized on the generalization with algebraic symbols which are

in generational and meta-level activities, although the study of

transformational activities has yet to be explored.15

The authors would like to know the implication of developing

algebraic thinking through technology concurrently the role of a

teacher. Although there are many benefits of using computers,

teachers also offer many advantages as well.

7. CONCLUSIONNowadays, computers are available to be used in the learning

environment.15 ICT tools allow students to use different strate-

gies through a stepwise approach with procedural skills as well

to enhance their conceptual understanding.6 Therefore, we can

conclude that the necessity of ICT tools in learning mathemat-

ics, particularly in developing algebraic thinking. However, the

role of teacher is still needed even though the ICT tools are

widely used. The teacher also should complete themselves with

the knowledge of content, ICT skills, and skills of questioning.

References and Notes1. O. S. Abonyi and I. Nweke, Journal of Education and Practice 5, 31 (2014).2. O. A. Alghtani and N. A. Abdulhamied, Procedia—Social and Behavioral

Sciences 8, 256 (2010).

3. S. Baltaci and A. Yildiz, Cypriot Journal of Educational Sciences 10, 12(2015).

4. R. Banerjee, Contemporary Education Dialogue 8, 137 (2011).5. L. Bills, K. Wilson, and J. Ainley, Research in Mathematics Education 7, 67

(2010).6. C. Bokhove and P. Drijvers, Computers and Education 58, 197 (2012).7. G. Booker and W. Windsor, Procedia—Social and Behavioral Sciences 8, 411

(2010).8. B. Dougherty, D. P. Bryant, B. R. Bryant, R. L. Darrough, and K. H.

Pfannenstiel, Intervention in School and Clinic 1 (2014), Doi: 10.1177/1053451214560892.

9. S. Gutiérrez, M. Mavrikis, and D. Pearce, A learning environment for promot-ing structured algebraic thinking in children, Eighth IEEE International Con-ference on Advanced Learning Technologies (2008), pp. 74–76.

10. J. Hall and G. Chamblee, Computers in the Schools: Interdisciplinary Journalof Practice, Theory, and Applied Reseacrh 30, 12 (2013).

11. C. Kieran, The Mathematics Educator 8, 139 (2004).12. J. Lee and J. Pang, Journal of the Korea Society of Mathematical Education

16, 2070 (2012).13. B. Patton and E. D. L. Santos, International Journal of Instruction 5, 5

(2012).14. R. Pierce, L. Ball, and K. Stacey, International Journal of Science and

Mathematics Education 7, 1149 (2009).15. M. Tabach, R. Hershkowitz, and T. Dreyfus, ZDM-International Journal on

Mathematics Education 45, 377 (2013).16. W. Windsor, Algebraic ThinkingD: A problem solving approach, Proceedings

of the 33rd Annual Conference of the Mathematics Education Research Groupof Australasia (2008), pp. 665–672.

17. W. Windsor, APMC 16, 8 (2011).18. W. Windsor and S. Norton, Developing algebraic ThinkingD: Using a problem

solving approach in a primary school context, MathematicsD: Traditions and[New]Practices (2011), pp. 813–820.

19. M. Zeller and B. Barzel, ZDM—International Journal on MathematicsEducation 42, 775 (2010).


3384




Vol. 21, 3389–3391, 2015

Significance of Preparedness in Flipped Classroom

Azlina A. Rahman∗, Baharuddin Aris, Mohd Shafie Rosli, Hasnah Mohamed,Zaleha Abdullah, and Norasykin Mohd Zaid

Department of Educational Science, Mathematics and Creative Multimedia, Faculty of Education,Universiti Teknologi Malaysia, Johor Bahru 81310, Malaysia

Flipped classroom is the latest pedagogy that has grown across multi-discipline and age levels which effec-tiveness has been proven empirically. The flipped classroom approach allows the students to review the topicsgiven prior to learning them in the classroom and apply the knowledge gained practically via in-class activities.Therefore, students are given more opportunities to apply the knowledge they have learned into the real lifesituation in collaborative learning environment. Furthermore, flipped classroom shows most impactful effects onincreasing learning interactions, improving students’ achievement and boosting critical thinking. Studies haveindicated that flipped classroom could also stimulate students’ interest and could even improve their attitudestowards school. Students are able to receive a personalized education to suit their learning style while syllabuscould be covered before time. However, the key feature of successful criteria in flipped classroom is students’preparedness. Very few of the reviewed articles emphasizes on this critical aspect. The need for the students tobe prepared prior to the teaching and learning process plays an important role to make this approach successfuland meaningful. This is because if the students come to the class unprepared, they will give a blank look andwill not get involved in the classroom. The school is a place to improve working improvement and to producestudents with maximum academic growth. While the flipped classroom approach has been seen successfulfrom the perspectives of both students and teachers, the authors notice that, there is still room for improvementin some areas. The authors choose to redesign the approach by factoring in the preparedness aspect. Thispaper summarizes the importance of preparedness based on limited past researches and also presents somepossible ways of redesigning the prior learning process in flipped classroom for secondary education.

Keywords: Flipped Classroom, Preparedness, Secondary Education, Critical Thinking.

1. INTRODUCTIONFlipped classroom is a pedagogical method that has started gain-

ing a place and attention in the world of education. Beginning

in the year 2000, the flipped classroom was first introduced by

Ref. [1] after realizing that the traditional teaching was nothing

more than copying lectures notes.

This gave rise to one-way interactions and students only had

their understanding tested through assessment and examinations.

As a result, lecturers could not rectify the students’ under-

standing right from the beginning of the lesson. The application

of the flipped classroom enables lecturer to identify students’

conceptual misunderstanding early, via two-way lecturer-students

interactions during the lectures.2–4

The style of teaching which is different from students’ learn-

ing style also tends to contribute to students’ misunderstand-

ing and this could eventually lead them to frustrations with the

learning experience especially when it involves with difficult

subjects. Taking the problem as a starting point,5 has applied


a concept similar with the flipped classroom, termed inverted

classroom.5 Realized that the incorporation of the students’ learn-

ing style and teachers’ learning styles improved the students’

performance and reduced the number of drop-outs. From the

cognitive aspects, previous studies also found the flipped class-

room capable of improving students’ performance and thinking

skills.6–10 In addition, the flipped classroom also helps in terms

of positive behavioral change, such as increasing students’ moti-

vation and reducing truancy.6�11 The use of technology in the

flipped classroom also helps students master their learning from

various aspects.

Studies related to the flipped classroom at the school level

began to be implemented widely after two school teachers in the

United States implemented the method on a group of high school

students learning chemistry.12 It was to be the starting point, after

which there were other researchers who applied the flipped class-

room approach for secondary school students.13 The findings of

the flipped classroom studies in secondary schools also showed

a positive impact on students’ performance14 and the curriculum

development.15 Generally the flipped classroom has a positive



impact on students’ performance, both at the tertiary9�16–23 and

the school levels.12�13�15�25�26

2. THE KEY FEATURE IN FLIPPED

CLASSROOMAlthough there have been many studies regarding the flipped

classroom in various dimensions and disciplines, those which

emphasize on students’ preparedness for the flipped classroom are

still lacking. The authors see that the key feature in the flipped

classroom is preparedness because if students are not prepared in

the first phase of the flipped classroom learning process, flipped

classroom learning objectives will not be achieved. Taking into

account some aspects to fulfill learning requirements of the stu-

dent, especially those in the secondary schools, the authors pro-

pose preparedness as one of the initiatives to be made available

in the flipped classroom learning process. Preparedness refers to

the stage of readiness of the students in the flipped classroom

approach.

The original flipped classroom learning process comprises two

learning phases.1 The first learning phase is a self-paced learning,

where teachers provide assignments for students to read about

what will be studied in the next lesson. Hence, during the flipped

classroom second phase session, the teachers could allocate more

time and provide more opportunities for their students. Interac-

tions between students and teachers could be improved. There-

fore, the teachers would find it easier to identify any students’

misunderstanding early. The students also learn from the teachers

and their peers who are more competent to promote the devel-

opment of knowledge and soft skills of the students. There are

previous studies26 which have made modifications on the flipped

classroom as shown in Figure 1.

Summarize the harvestpropose confusion

Learning by one selfSelf paced

Exhibit communicationResearching cooperatively

Scientific experimentAccomplish Homework

Layout preview

Out

of c

lass

In th

e cl

assr

oom

Fig. 1. Teaching structure.26

Redesign Phase: Prior to teaching andlearning process

Option 1:Students are given questions about what

is read in Phase 1. This may be done either online or offlineor both.

Option 2:Students perform instantaneous activitiesrelated to what has been learnt. It couldbe as in the written form such as quizzes

and education game.

Option 3:Each student is to ask questions to the

teachers or peers.

Phase 2

Activities in the classroom such as groupdiscussion, demonstrations, lab activities

and debate.

Phase 1

Students are exposed to read or referencematerials through the use of digital

technologies such as video and the InternetO

ut o

f cla

ssIn

the

clas

sroo

m

Fig. 2. Redesign stage of flipped classroom preparedness.

According to Figure 1, the flipped classroom learning process

is divided into two phases which are learning in the class-

room and learning outside the classroom. Learning in the class-

room means self paced learning while learning outside the

classroom consists of hands-on learning. However, the teaching

structure shown in Figure 1 is feasible at the tertiary level26 has

developed the activities as portrayed in the dashed box. In the

context of this study, the authors see the potentials if the steps

in the dashed box could be applied in various levels, particularly

at the school level. The authors feel that there is some looseness

in Figure 1, especially in the dashed box, namely, what if the

students do not make the task as directed by the teachers. If the

students are not ready, they will come to class unprepared with

a blank look. This situation will potentially become worse as the

students are not able to fully absorb the lesson which will take

place. It is to cater for this need that, the author has drawn up a

guideline in the flipped classroom learning process, as depicted in

Figure 2. These guideline is strongly suggested by Refs. [27, 28]

that the teacher or instructor must also develop activities and/or

pretest to ensure that students are prepared for class.

In reference to Figure 2, the original version of flipped class-

room consists of two phases, namely Phase 1 (out of class) and

Phase 2 (in the classroom) as shown in Figure 1. As the author

3390


sees the importance for preparedness in the flipped classroom

to be emphasized, the author has redesigned a phase marked in

the dashed box in between Phase 1 and Phase 2 in Figure 2.

Although there are researchers21�29 who are concerned about the

preparedness such with emphasis on online learning using Intelli-

gent Tutoring System as one of the initiatives to ensure that each

student is interactive with what has been learnt, the implementa-

tion of such an approach is limited only to the university students.

At the school level, only researchers12 have made it compulsory

for each student to submit questions before the class starts to

ensure that they are ready for the assignment to be given. It is

obvious that, studies which emphasize on the preparedness aspect

of the flipped classroom research are very limited. Findings in a

research by Refs. [27–29] prove that students’ preparedness with

regard to digital materials or assignment given prior to teaching

and learning results in active learning, compared to students who

are not prepared with the provided materials.

The authors see redesigning the flipped classroom approach

as a measure that needs to be done so that students are more

prepared for the assigned learning materials. Using the mea-

sures provided, it will assist teachers to ensure the success of

the flipped classroom approach. The redesigned method provides

teachers with the guidelines to implement the steps shown before

going to Phase 2 in Figure 2. Based on Figure 2, for schools

with adequate infrastructure facilities such as Internet and com-

puter hardware, Option 1 in the redesigned phase does not pose a

problem. However, for schools with problems in terms of infras-

tructure, they may opt for off line learning. For teachers who have

the time and creativity, they may capitalize on this opportunity

by giving written exercises such as quizzes and education games

as proposed in Option 2. The simplest approach is Option 3,

in which, the teachers require each student to ask questions and

the discussion sessions are continued in the following phase as

made by researchers.12

Comparing Figure 1 with Figure 2, especially in the dashed

boxes, the authors emphasize the preparedness aspect, without

which, the objectives of the flipped classroom approach are hard

to be achieved. This situation could occur not only at the school

level but also at the tertiary level. In the context of this study, the

authors chose an Option 1 to be applied outside the classroom

as a self-pace learning.

3. CONCLUSIONOverall, flipped classroom has a positive impact on students’

achievement. However, when viewed in terms of infrastructure,

it should be given a little touch to make it more feasible.

The authors see the need for this issue to be highlighted with

the redesigned approach with the introduction of preparedness

aspect, especially in the secondary education. Secondary school

students need to be kept motivated to make learning more inter-

esting and meaningful. The proposal by the authors are also

strongly suggested by Refs. [27–29] who also emphasized on the

importance of students’ preparedness before the flipped class-

room teaching and learning begins. Although the researcher sug-

gested a preparedness phase in applying the flipped classroom,

the use of online and offline technology also helps in improving

students’ achievement.12�14�20–26�30 However, the survey data can-

not be reported since the data collection process is still ongoing.

Acknowledgments: The authors would like to honour

the continuous support and encouragement given by Univer-

siti Teknologi Malaysia (UTM) and the Ministry of Educa-

tion Malaysia (MoEM) in making this research possible. This

academic work was supported by Fundamental Research Grant

Scheme (RJ130000.7816.4L088) initiated by MoEM.

References and Notes1. W. Baker, The ‘classroom flip’: Using web course management tools too

become the guide by the side, 11th International Conference on CollegeTeaching and Learning, Jacksonville, FL (2000).

2. B. Tucker, The flipped classroom, Education Next (2012), pp. 82–83.3. J. G. Jr, The Effects of a Flipped Classroom on Achievement and Student

Attitudes in Secondary Chemistry (2013).4. J. Chipps, The Effectiveness of Using Online Instructional Videos with Group

Problem Solving to Flip the Calculus Classroom (2013).5. M. J. Lage, G. J. Platt, M. Treglia, and J. Lage, J. Econ. Educ. 31, 30 (2000).6. C. F. Herreid and N. A. Schiller, J. Coll. Sci. Teach. 42 (2012).7. K. Lockwood, C. S. U. M. Bay, and R. Esselstein, The inverted classroom

and the CS curriculum, Proceeding of the 44th ACM Technical Symposiumon Computer Science Education, New York, USA (2013), pp. 113–118.

8. C. Demetry, Work in progress–An innovation merging classroom flip andteam-based learning, 40th IEEE Frontiers in Education Conference (FIE)(2010).

9. R. H. Rutherfoord and J. K. Rutherfoord, Flipping the classroom—Is it for you?14th annual ACM SIGITE Conference on Informtion Technology Education(2013), pp. 19–22.

10. N. Hamdan, P. McKnight, K. McKnight, and K. M. Arfstrom, Flip. Learn. Netw.(2013).

11. A. Butt, Bus. Educ. Accredit. 6, 33 (2014).12. J. Bergmann and A. Sams, Learn. Lead. with Technol. 36, 22 (2009).13. K. P. Fulton, Phi Delta Kappan J. Storage 94, 20 (2012).14. D. Siegle, Differentiating Instruction by Flipping the Classroom (2013), Vol. 37,

pp. 51–56.15. S. Flumerfelt and G. Green, Educ. Technol. Soc. 16, 356 (2013).16. N. K. Pang and K. T. Yap, The Flipped Classroom Experience, IEEE CSEE

and T 2014, Klagenfurt, Austria (2014), pp. 39–43.17. K. Peacock, Flipping Your Classroom?: A Ticket to Increased Classroom Col-

laboration?, Centre for Teaching and Learning University of Alberta (2013).18. D. N. Shimamoto, Implementing a flipped classroom?: An instructional mod-

ule, Technology, Colleges, and Community Worldwide Online Conference(2012).

19. B. B. Stone, Flip your classroom to increase active learning and studentengagement, 28th Annual Conference on Distance Teaching and Learning(2012), pp. 1–5.

20. A. Steed, Proquest 16, 9 (2012).21. J. F. Strayer, Learn. Environ. Res. 15, 171 (2012).22. N. Warter-Perez and J. Dong, Flipping the classroom?: How to embed inquiry

and design projects into a digital engineering lecture, Proceedings of the 2012ASEE PSW Section Conference (2012).

23. Z. Zhang, Construction of online course based on flipped classroom model(FCM) concept, 2nd International Conference on Information, Electronics andComputer (ICIEAC) 2014, September (2013), pp. 157–160.

24. D. Siegle, Gift. Child Today 37, 51 (2013).25. K. R. Clark, Examining the Effect of the Flipped Classroom Model of Instruc-

tion on Student Engagement and Performance in the Secondary Mathematicsclassroom: An Action Research Study (2013).

26. Y. Jiugen, X. Ruonan, and Z. Wenting, Essence of flipped classroom teachingmodel and influence on traditional teaching, IEEE Workshop on Electronic,Computer and Applications (2014), pp. 362–365.

27. J. A. Day and J. D. Foley, IEEE Trans. Educ. 49, 420 (2006).28. S. Kellogg, Developing online materials to facilitate an inverted class-

room approach, 39th ASEE/IEEE Frontiers in Education Conference (2009),pp. 1–6.

29. S. Zappe, R. Leicht, J. Messner, T. Litzinger, and H. W. Lee, Flipping the class-room to explore active learning in a large undergraduate course, AmericanSociety for Engineering Education (2009).

30. A. A. Rahman, N. Mohd Zaid, Z. Abdullah, H. Mohamed, and B. Aris,Emerging project based learning in flipped classroom, The 3rd InternationalConference of Information and Communication Technology (ICoICT) (2015),In revised.


3391




Vol. 21, 3392–3395, 2015

Building a Data Mart Using Single Dimensional

Data Store Architecture with Student Subject: Case

Study at Muhammadiyah University of Yogyakarta

Fajar Rianda∗, Asroni, and Ronald Adrian

Muhammadiyah University of Yogyakarta, Indonesia

Muhammadiyah University of Yogyakarta have an academic information system. This system is an Online Trans-action Processing (OLTP). System have data which related with student and can used for accreditation. But,system have not capability to give information quickly and accurately. Data warehouse is a system which havecapability to give information quickly and accurately. With this ability to easily get related information students.This research using Single Dimensional Data Store (DDS) architecture. The goal research is build student datamart on data warehouse of Muhammadiyah University of Yogyakarta using single DDS architecture and result-ing information quickly and accurately for supporting borang accreditation. Data mart has been built give easeto get information about student and the information can display in tabular.

Keywords: Data Warehouse, Data Mart, Borang Accreditation, Student, Single DDS.

1. INTRODUCTION1.1. Background

Every college have to accredited by Badan Akreditasi Nasional–

Perguruan Tinggi (BAN-PT). Accreditation is used to determine

eligibility and quality of college. BAN-PT will accrediting using

borang accreditation form.

Muhammadiyah University of Yogyakarta (UMY) as one of

the college who have to fill in the accreditation. UMY have an

academic information system. This system is an Online Trans-

action Processing (OLTP). System have data which related with

student and can used for accreditation. But, system have not

capability to give information quickly and accurately.

The solution for that case is data warehouse. Data warehouse

is a system which have capability to give information quickly

and accurately. With this ability to easily get related information

students.

1.2. Formulation of the Problem

Formulation of the problem based on the background above are

how to building data mart on data warehouse with single dimen-

sional data store architecture.

1.3. Goal

The goal of this research is to build student data mart at of

Muhammadiyah University of Yogyakarta using single DDS


architecture and resulting information quickly and accurately for

supporting borang accreditation.

2. THEORITICAL2.1. Literature Review

The Research related about data warehouse has been done several

times. Some research as a reference for this research is:

• Windarto has done research that entitle Pemanfaatan Data

Warehouse sebagai sarana Penunjang Penyusunan Borang

Akreditasi Standar 3 pada Fakultas Teknologi Informasi Univer-

sitas Budi Luhur.1 His conclusion research is using data ware-

house give ease to get information more effective and increase

data security.

• Mukhlis Febriady and Bayu Adhi Tama have done research

that entitle Rancang Bangun Data Warehouse untuk Menunjan

Evalusai Akademik di Fakultas.2 Their research resulting inte-

grated database and make report faster.

• Other research from Armadiyah Amborowati that entitle Per-

ancangan dan Pembuatan Data Warehouse pada Perpustakaan

STMIK AMIKOM Yogyakarta.3 Her research made load process

periodic automatically and help administrator.

2.2. Data Warehouse

Data Warehouse is a subject-oriented, integrated, time-variant,

nonupdatable collection of data used in support of management



Fig. 1. Research procedure.

decision-making processes and business intelligence.4 The mean-

ing of each of the key terms in this definition follows:5

• Subject-oriented, a data warehouse is organized around the

key subjects.5 Major subjects may include students, patients, and

products.5

• Integrated, the source data from different source systems. The

source data usually inconsistent. The integrated source system

must be made consistent.6

• Time-Variant, Data in the warehouse is only accurate and valid

at some point in time or over some time interval.6

• Nonupdatable, Data in the data warehouse are loaded and

refreshed from operational system, but cannot be updated by end

users.5

2.3. Online Transaction Processing (OLTP)

Online Transaction Processing is a system whose main purpose

is to capture and store the business transactions.7 OLTP system

contain the data that will load into the data warehouse.7

2.4. Difference of OLTP and Data Warehouse

See Table I.

2.5. Data Mart

Data mart is a subset of a data warehouse that supports the

requirements of a particular department or business function.6

Fig. 2. Single DDS.

Fig. 3. Star schema.

Table I. Difference of OLTP and Data Warehouse.6

OLTP Data warehouse

Holds current data Hold historical dataData is dynamic Data is largely staticTransaction-driven Analysis-drivenApplication-oriented Subject-orientedSupports day-to-day decision Supports strategic decisions

2.6. Extract, Transform, and Load (ETL)

ETL is a system that has the capability to connect to the source

systems, read the data, transform data, and load it into a target

system.7

2.7. Dimensional Data Store

Dimensional Data Store is a database that stores the data ware-

house data in a different format than OLTP.7 DDS is a better

format to store data in the warehouse for the purpose of querying,

analyzing data and gives better query performance.7

2.8. Dimensional Modelling (Star Schema)

A Star schema is a simple database design in which dimensional

data are separated from fact or event data.5 A star schema consist

of two types of tables: fact tables and dimension tables. Star

schema is simpler than other schema, and making easier for ETL

process data into DDS.7

2.8.1. Single Dimensional Data Store (DDS)

Architecture

In Single DDS, consist of only two data stores: stage and DDS.

Single DDS architecture use two data store, which are:7

(a) A stage, is a place where you store the data you extracted

from the store system temporarily, before processing it further or

load into to other data store.

(b) Dimensional Data Store (DDS), is a user-facing data store,

where the data is arranged in dimensional format for purpose of

supporting analytical queries.7

Single DDS is simpler because the data from the stage is

loaded straight into the dimensional data store, without going to

any kind of normalized store first.7

Table II. Name changes from source store to stage.

Source system Stage data store

dbo.MAHASISWA dbo.stage.mahasiswadbo.STATUS_TERDAFTAR dbo.stage.status_terdaftardbo.STATUS_TRANSFER dbo.stage.status_transferdbo.THAJARAN dbo.stage.thnajarandbo.DEPARTMENT dbo.stage.departmentdbo.FACULTY dbo.stage.facultydbo.ANGKATAN dbo.stage.angkatandbo.CLASS_PROGRAM dbo.stage.class_program

Table III. Name changes from stage store to DDS.

Stage data store Dimensional data store

dbo.stage.mahasiswa dbo.dim_mahasiswadbo.stage.status_terdaftar dbo.dim_status_terdaftardbo.stage.status_transfer dbo.dim_status_transferdbo.stage.thnajaran dbo.dim_thnajarandbo.stage.angkatan dbo.dim_angkatandbo.stage.class_program dbo.dim_class_program

3393


Fig. 4. Source system.

3. METHOD

In this research there are some steps. The steps are display in

Figure 1.

3.1. Requirement

Requirement is first step in this research. Requirement will deter-

mined as basis for the building a data mart.

3.2. Building

In this research using single DDS architecture. In single DDS

there are some step to build. Single DDS architecture display in

Figure 2.7

Fig. 5. Star schema.

In Figure 2, the first step is determine source system which

will use for building data mart. After source system have deter-

mined, the next step is ETL process for load data from source

system into stage data store. After data have been loaded in stage

data store, the next step is ETL process again and data quality

(DQ). In second ETL process data will stored in dimensional

data store with star schema structure. Star schema display in

Figure 3.

3.3. Testing

There are two testing will using in this research, which are:

• ETL testing, in this testing make sure that appropriate changes

in the source system are loaded correctly into data mart.

3394


Fig. 6. Database architecture.

Table IV. ETL testing.

Data store

Year Stage store DDS

2014 6380 63802013 7003 70032012 5391 53912011 3912 39122010 2714 27142009 2174 21742008 1878 18782007 1916 19162006 1615 16152005 2150 21502004 2467 24672003 2515 25152002 3174 31742001 3061 30612000 2946 2946

• Functional testing, in the functional testing verify make sure

all requirements are fulfilled by data mart.

4. RESULT AND EXPLANATION4.1. Requirement

The requirement data in this research obtained from borang

accreditation document standard 3.

4.2. Building

4.2.1. Source System

The source system that used is master tabel database. Data that

taken from the students of 2000 to 2014. In master tabel database

there are some tables that used for this research. The following

tables are:

• dbo.MAHASISWA

• dbo.STATUS_TERDAFTAR

• dbo.STATUS_TRANSFER

• dbo.THAJARAN

• dbo.FACULTY

• dbo.DEPARTMENT

• dbo.ANGKATAN

• dbo.CLASS_PROGRAM

as for the figure relation between tables above as follows.

4.2.2. Extract, Transform, and Load (ETL)

In first ETL process, source system will stored in stage data store

and change the name of tables. Table II displaying tables name

changes in stage data store.

Table V. Student count.

Student count

Academic periodic Regular non transfer Transfer

TS-4 2587 0TS-3 3719 0TS-2 5211 0TS-1 6775 0TS 6072 0

After data have stored in stage data store, the next step is ETL

process. In second ETL process, data will stored from stage to

dimension data store. Beside store data, in this step will do data

quality. Data quality make sure data that stored in dimensional

data store are not dirty data or noise. Dirty data or noise can

include null data, duplicate data. In this step, name of tables

will change. For dbo.stage.faculty table and dbo.stage.department

table merged and became dbo.stage.dim_faculty_department.

Table III displaying tables name changes in DDS:

That tables are dimension table. In DDS, there are one fact

table. The table is dbo.fact_jumlah_ mahasiswa. Dimension table

and fact table made with star schema structure. Figure star

schema display in Figure 5.

As the result of database architecture that have made from

source system to dimensional data store is shown in the following

figure.

4.3. Testing

4.3.1. ETL Testing

In this testing will make sure that data in ETL process have

stored to data store. The results of testing display in Table IV.

4.3.2. Functional Testing

This testing make sure that data mart have capability to fulfill

requirement of borang accreditation document. Table V display

result of testing.

5. CONCLUSIONThe conclusion that can taken from this research is building

data mart give ease to get information about student, can used

by Muhammadiyah University of Yogyakarta to fill in borang

accreditation and the information can display in tabular.

References and Notes1. Windarto, Telematika MKOM 3 (2011).2. M. Febriady and B. A. Tama, Rancang bangun data warehouse untuk menun-

jang evaluasi akademik di fakultas, KNTIA, Palembang (2011).3. A. Amborowati, Perancangan dan pembuatan data warehouse pada per-

pustakaan stmik amikom yogyakarta, Seminar Nasional Aplikasi Sains DanTeknologi (2008).

4. W. H. Inmon, Building the Data Warehouse, Fourth edn., John Wiley and Sons(2005).

5. J. A. Hoffer, M. B. Prescott, and F. R. McFaden, Modern Database Manage-ment, Eightth edn., Pearson Education Inc. (2007).

6. T. M. Connolly and C. E. Begg, Database System A Practical Approach toDesign, Implementation, and Management, Fourth edn., edited by AddisonWeslet, United States of America (2005).

7. V. Rainardi, Building a Data Warehouse: With Example in SQL Server (2008).


3395




Vol. 21, 3396–3399, 2015

Contributing Factors of Online Brand

Trust in Airline Industry

Nur Atika Jamuary1, Mohd Shoki Md Ariff1�∗, Hayati Jamaludin1, Khalid Ismail2,Nawawi Ishak3, and Mohd Sawal Abong3

1Faculty of Management, Universiti Teknologi Malaysia, 81300 Skudai, Malaysia2Universiti Pendidikan Sultan Idris, 35900 Tg. Malim, Malaysia

3Lembaga Tabung Haji Malaysia

The advancement of online channel has point out the importance of online brand trust in order to ensuresustainability of airline business. Online trust plays an important role in creating satisfied and expected outcomesbetween online buyers and airline companies, and the companies need to address factors contributing to theonline brand trust. The increase in international students’ intake in Malaysia provides huge opportunity foronline airline tickets service providers because they are flying to and from Malaysia on a continuous basis.However, studies on online brand trust are still limited in airline industry especially among international studentspursuing their study in Malaysia. A conceptual model of online brand trust featuring six factors—word of mouth,security/privacy, perceived risk, good online experience, brand reputation and perceived quality of information—was developed based on studies of brand trust to determine factors contributing to the online brand trust of aMalaysian-based airline company. Analysis of data was performed based on the 276 respondents, which werecollected using a convenience sampling method from international students in a public university of Malaysia.The result indicated that there are five factors contributing to the online brand trust in the airline industry—word of mouth, security/privacy, perceived risk, good online experience/brand reputation and perceived qualityof information. Good online experience/brand reputation appeared to have stronger effect on online brand trust.Theoretical and practical implications of this study were discussed in understanding online brand trust involvinginternational students and ways to enhance their trust towards the airline companies.

Keywords: Brand Trust, Online Brand Trust, Airline Industry.

1. INTRODUCTIONOnline brand trust in the e-commerce study is very important

because it facilitate trust-related behaviors between online buyers

and online sellers, such as making purchases and stay loyal to the

sellers.1 From online buyers’ point of view, online brand trust is

very crucial because it helps them reducing risks and uncertainty

related to the online purchasing. For online sellers, in order to

build a sustainable relationship with online customers, the lack

of brand trust would eliminate commitment of customers towards

their brand. In airline industry, in which most of consumers buy

airline ticket online, understanding how airline carriers can gain

the trust of customers is highly needed.

With the rapid advancement of information and communica-

tion technology, the Internet has turned into an important tool in

business, especially in the area of online reservation in the air-

line industry.2�3 The rate of purchasing the airline e-reservation

ticket globally has expanded by 8% to 32% from year 2008 to

2010.4 In Malaysia, the number of internet users is accounted


around 15.355 million users (CIA, 2011) and 15.4% of the inter-

net users in Malaysia utilized the online shopping, and 54.7% of

this involved in purchasing of airline tickets online.5 The increase

purchase of airline tickets online highlights the importance of

understanding factors associating with the online brand trust in

airline industry. Most previous studies of brand trust addressed

these factors in conventional air ticket purchasing that may not

be applicable in online purchase of airline tickets. In addition,

factors that affect online brand trust in airline industry could be

different from those that affect it in other online industries.

Review of previous studies related to factors affecting online

brand trust in airline industry from the international students’

perspective has not been addressed. Airline companies spent con-

siderable amount of money to create trust and loyalty among

the consumers, thus, there is an increasing need to understand

online brand trust, and therefore gain the greatest return on their

investment.6 International students purchase online airline tickets

on a regular basis throughout their studies. Furthermore, given

that the international students are considerably high in the pur-

chase of airline ticket online, revealing factors affecting their

online brand trust would be interesting to research. Further, this



study is important for practitioners and academics because much

of the work on trust has been theoretical rather than empirical

and there has even been empirical work on trust online.1�7

2. LITERATURE REVIEWBrand trust refers to psychological elements that illustrate the

whole assumption related to credibility and integrity that the cus-

tomer would describe to that particular brand.8 It is a feeling

of security that consumers have in their dealings with a brand,

which it is based on the perceptions that the brand is depend-

able and liable for their interests.9 In internet setting, such as

purchasing of airline ticket online, online buyers who had trust

in a brand are willing to share information and their experiences

about the brand10 because they feel safe and comfortable doing

online transaction with regard to the brand. Thus, online brand

trust is a critical factor in motivating purchases over the inter-

net and therefore, it is important for online company to win and

build customers trust in order to survive in a very competitive

marketplace.11 In internet-based transaction, online buyers who

perceived high level of trust on a particular brand seems to be

satisfied with the brand12 and may turn into loyal consumers.13

In the classic Theory of Commitment-Trust, Morgan and

Hunt14 theorize that successful relationship marketing requires

relationship of commitment and trust between seller and buyers.

In an online environment, this relationship could develop online

buyers’ brand trust towards the online sellers. This psychological

state of mind illustrates credibility and integrity that an online

buyer would describe an online brand;8 a perception that the

online brand is dependable and liable for his or her interest.9

The presence of relationship commitment and trust is principal

to successful relationship marketing instead of company’s power

and its ability to persuade others. Trust is essential to relational

exchange that Spekman15 claims it to be the basis of the strate-

gic partnership. In online selling, trust is crucial because it leads

online buyers to form a feeling of safe and secure and this will

reduce their perceived risk. Trust are extremely valued by both

online buyers and online sellers, and for buyers, this will facili-

tate their interest to continuously engage in online purchasing.

Based on the review of the contributing factors of brand

trust in online setting, it can be summarized that there are

six factors affecting online brand trust—word of mouth,13�16–18

security/privacy and good online experience,6�13�16–18 perceived

risk,17�18 perceived quality of information6�13�17�18 and brand

reputation.13�17�18 These six factors (independent variables) are

hypothesized to have significant effect on online brand trust

(dependent variable), as shown in the Figure 1—The Conceptual

Model of the Study. This model is based on the Commitment-

Trust Theory and past researches on contributing factors of brand

Word of Mouth

OnlineBrand Trust

Security/Privacy

Perceived Risk

Good Online Experience

Quality of Information

Brand Reputation

Fig. 1. The conceptual model of the study.

trust in online setting. Other factors that contribute to online

brand trust are website design and navigation6�18 and advertising

and testimonial.6 The first six factors of online brand trust are

related to the individual online buyer; therefore these two factors

are excluded from this study.

Studies in brand trust placed a very high level of concern

about security and privacy factors. In fact, both factors have

always been important, they take on a wider significance on

online channel.6�11�13�17–20 Security is defined as the belief of the

customer that the web or online channel is secured to convey

the private information.21 Besides, privacy refers to individual

that has the ability to control his personal information which

is acquired and used by other party.22 Ha13 stated that security

and privacy influenced customers’ trust towards a brand whereby

customers concern on higher security and privacy feeling with a

higher level of brand trust. Most of the past researchers agreed

that safety and security have always been important element in

online business in order to gain customer’s trust.

In online purchasing, consumer perceived risk will influence

the development of brand trust because it leads to the feeling of

secure and reduce uncertainty about possible negative outcomes

of an online transaction. Perceived risk refers to a customer’s

belief on the uncertainty results that might happen during online

transaction because no matter what happens customers always

want to avoid mistakes rather than to maximize the risk while

committing in online purchasing.23 In order to avoid perceived

risk, customers usually use several strategies which include brand

loyalty, brand trust, store image or word of mouth.24 In a high

perceived risk situation, they always compare alternatives and

ask friends and relatives for assistance to reduce the feeling of

uncertainty.25 Thus, perceived risk is a contributing factor of

online brand trust. In fact, perceived risk is higher in an online

environment as compared to an offline environment.25

The online purchasing involve experience of searching, brows-

ing, finding, selecting, comparing and evaluating information as

well as interacting and transacting with the online company. Most

researchers found that good online experience is one the most

important factors of online brand trust.6�11�17�18 Based on Ha,13 a

good online experience would affect positive WOM, brand loy-

alty, brand trust and online repurchase intention. Consumers’ pre-

vious online shopping experience with online company may help

them to gain more trust in shopping via the internet site.26

Word of mouth (WOM) communication has a greater impact

on customer trust on online business, as it creates viral and

convey the message faster on the internet compared to offline

environment.13 WOM refers to the non-formal communication in

order for online buyers to evaluate the goods and services offered

by online sellers.27 The quality of word of mouth is correlated

to online shopping indicating WOM positive effects on online

brand trust.28

Brand reputation gives a product its core identity and cannot

be changed easily.29 Customers are aware that most of preferred

brand that have good reputation would provide trust, familiarity

and comfort for them.13 Customer would perceive high level of

trust from the company that has good brand reputation.17 Thus,

brand reputation is important in establishing online buyers’ trust

towards an online brand.

In online purchasing, a buyer would evaluate the web sites

according to individualized information needs whereby the more

useful functions or information that web sites provide, the higher

3397


the online initial trust.30 Ha13 believes that online company

should focus on the quality of information rather than the quan-

tity of information and try to offer customized information on the

web to provide a good online experience for customer through

online channel. In order to enhance early perceptions of trust, the

web site ought to be refined by informative and useful content,

wisely selected functionalities and focusing on target customers’

requirements.31 Therefore, quality of information in online sell-

ers’ websites is crucial in building up online buyers’ trust.

3. METHODOLOGYThe conceptual model of this study is based on the Theory

of Commitment-Trust.14 Based on review of literature on fac-

tors contributing to online brand trust, six factors were cho-

sen to determine online brand trust for an airline company. A

questionnaire was developed to measure online brand trust of

international students who are pursuing their studies in a Pub-

lic University in Malaysia. The 27 items developed measures the

following:

• four items constructed for WOM were adapted from

Ha;13�17�18 five items for security/privacy;6�17�18 four items

for perceived risk;17�18 and three items each for good

online experience, perceived quality of information and brand

reputation.13�17�18

• five items of online brand trust was adopted from Alam and

Mohd Yasin.17

The scale of items was measured based on five points Lik-

ert Scale ranging from 1-strongly disagree to 5-strongly agree.

The population of this study is all international students in a

public university who have experience purchasing airline ticket

online from the airline website. 270 sample sizes are required

to achieve 5% margin error.32 The sampling technique used

was convenience sampling because the total number of students

who purchased airline ticket online is unknown. Students were

approached and ensured they have experience with the purchase

prior to completing the questionnaire. The total number of com-

pleted questionnaires received was 276, i.e., 92% response rate.

For instrument assessment, the normality test, linearity and

multicollinearity tests were performed to make sure the underly-

ing assumptions of correlation and regression analysis were met.

The KMO score for the six independent variables is 0.848, sup-

ported by Barlett’s test of Sphericity with significant values of

.000. The Total Variance Explained of Exploratory Factor Anal-

ysis (EFA) is 76.339% with five factors with eigenvalues greater

than 1. The rotated component matrix had produced five compo-

nents and 19 items in each component were retained since all the

factor loadings were above 0.5. However, three items for WOM,

good online experience and brand reputation have been deleted

since their factors loading of >0.5 appeared in more than one

component. Four items (two items of good online experience and

two from brand reputation) were loaded in the same component

and they were renaming as online experience/brand reputation.

The other items for each factor were retained in the same factor

as they were loaded in the same component.

In the second round of EFA, the KMO measure of sampling

adequacy for the five independent variables is 0.834, supported

by Bartlett’s test of Sphericity with significant value of 0.000.

Five factors with eigenvalues greater than 1 had been extracted

with the Total Variance Explained of 76.177%. The rotated com-

ponent matrix had produced five components and all the items in

each component were retained since all the factor loadings were

above 0.5. Therefore, based on the EFA results, the five inde-

pendent variables are word of mouth, security/privacy, perceived

risk, online experience/brand reputation and perceived quality of

information.

For online brand trust, the KMO score is 0.814, supported with

Barlett’s test of Sphericity with significant values of .000. One

factor with eigenvalues greater than 1 had been extracted in the

EFA with the cumulative percentage of 67.575%. All the five

items were retained as the loading factors were >0.5. For reliabil-

ity analysis, the Cronbach’s Alpha for WOM is 0.730, 0.929 for

privacy/security, 0.917 (Perceived risk), 0.859 for online experi-

ence/brand reputation, quality of information (0.771) and online

brand trust 0.874. Thus, all key constructs were statistically reli-

able and valid for further analysis.

4. RESULT AND ANALYSISMultiple Regressions analysis was performed to determine the

effect of all the five factors on online brand trust. The conceptual

model does not consist of unobservable latent variable and not

too complex that requires structural equation modelling analysis.

The result of this analysis, as shown in Table I, indicated that

the five factors significantly contributed to the online brand trust

of the respondents. Specifically, the respondents’ WOM (��142,

p < �005), perceived Security/Privacy (��153, p < �005), Per-

ceived risk (��150, p < �001), online experience/brand reputa-

tion (��466, p < �001) and perceived Information quality (��309,

p < �001) contributed positively to the online brand trust. Thus,

in airline industry, in which most buyers purchase airline ticket

online, these are factors that contribute significantly to their trust

towards the airline company.

5. DISCUSSIONSThe main finding of this study indicated that, in airline industry,

the contributing factors to online brand trust are word of mouth,

security/privacy, perceived risk, online experience/brand reputa-

tion and perceived quality of information. It shows that when

buyers purchase airline air ticket online, their trust towards the

brand is influenced by word of mouth communication among

them, how they perceived security/privacy and risk of the air-

line website, their good online experience with the site as well

as the brand reputation perceived. Airline ticket purchase infor-

mation describing how the process is organized and information

Table I. Multiple Regression of online brand trust.

Parameter B SE � t Sig. VIF

Constant �579 .239 2.418 0�0016Word of mouth 0�163 0.050 0�142∗ 3.244 0�001 1.498Security/Privacy 0�138 0.044 0�153∗ 3.155 0�002 1.839Perceived risk 0�121 0.032 0�150∗∗ 3.724 0�000 1.257Online experience/ 0�431 0.046 0�466∗∗ 9.351 0�000 1.939brand reputation

Information quality 0�341 0.046 0�309∗∗ 7.425 0�000 1.348

F 101.864 R 0.732 R2 0.808

Note: ∗p < 0�05; ∗∗p < 0�01.

3398


related to the payment, available service, and et cetera is also a

contributing factor to online brand trust.

Word of mouth was found to have positive and significant

effect on online brand trust. This result is consistent with Ha,13

Kim and Song28 and Alam and Mohd Yasin17 who found pos-

itive and significant relationship between word of mouth and

online brand trust. Online buyers’ word mouth communication

is considered as important factor that would likely to influence

their online buying behavior in airline services. It shows that the

increasing of word of mouth that the brand has, the higher the

level of trust that customer has toward that brand.

Security was positively and significantly affected online brand

trust, which is consistent with the works of Ha,13 Srinivasan16

and Alam and Mohd Yasin.17 When privacy rules are clearly

defined in the airline company’s website and customers do not

have any worries on providing of their personal information, it

indicate that they are more likely to trust the airline brand. This

indicated that the more customers feel secure about online brand

and have confidence in providing their personal information, the

more they trust that online brand.

Furthermore, perceived risk was also found to have positive

effect on online brand trust. However, this result contradicts with

those of previous studies done by Alam and Mohd Yasin17 and

Mohammadian and Ghanbar.18 Previous studies have found that

perceived risk has no significant influence on online brand trust.

This contradicting result might be based on the fact that many of

the international students’ perception of risk regarding of buying

online ticket from the airline company’s website may not have

impact on their trust of that online brand.

The results had shown that good online experience/brand rep-

utation had positive effects on online brand trust. This result

was also observed in the previous research carried out by Ha,13

Ruparelia et al.,6 Alam and Mohd Yasin17 and Mohammadian

and Ghanbar.18 Online buyers are seemed to have trusted on the

airline brand and they are more likely to repurchase from the air-

line website because they have a pleasant experience dealing with

the online purchase of air ticket. Therefore, good online experi-

ence/brand reputation is a factor contributing to higher trust in

online brand.

Previous studies conclude that the higher the quality of infor-

mation that online brand’s website offers, the higher the level of

brand trust the consumer gained.6�13�17�18 This study supports this

view in which perceived quality of information positively affect

online brand trust. It calls online companies to focus more on

providing information that can satisfy and fulfil their online cus-

tomers’ needs instead of providing unnecessary and unorganized

information.13

6. CONCLUSIONSThis study concluded that the contributing factors to online brand

trust in airline industry are word of mouth, security/privacy, per-

ceived risk, online experience/brand reputation and perceived

quality of information. These factors are very much related to

the perception of online buyers, and may not be sufficient to

determine brand trust in online setting. In the purchase of air-

line ticket, buyers may encounter difficulties related to the func-

tionality of the website. Therefore, factor of web design and

navigation6�18 should be considered in examining online brand

trust. Advertising and testimonial6 could also be used in the con-

ceptual model of online brand trust to comprehensively capture

all possible factors that influence online buyers to develop their

trust in online setting.

In this study, international students who are pursuing their

studies in Malaysia had expressed their views on factors con-

tributing to the online brand trust. Further study should be carried

out to generalize contributing factors of online brand trust in air-

line industry by extending participation of various segments of

online air ticket.

References and Notes1. D. H. McKnight, V. Choudhury, and C. Kacmar, Information System Research

13, 334 (2002).2. D. Buhalis and R. Law, Progress in information technology and tourism man-

agement: 20 years on and 10 years after the Internet—The state of eTourismresearch, Tourism Management (2008), Vol. 29, pp. 609–623.

3. E. Bigne, B. Hernandez, and L. Andreu, Journal of Air Transport Management16, 346 (2010).

4. C. W. Wang, Airline ticket E-Reservation: Adoption among Malaysians, Masterthesis, USM (2011).

5. Malaysian Communication and Multimedia Commission (2008),www.skmm.gov.

6. N. Ruparelia, L. White, and K. Hughes, Journal of Product and BrandManagement 19, 250 (2010).

7. S. L. Jarvenpaa, N. Tractinsky, and M. Vitale, Information Technology andManagement 1, 45 (2000).

8. D. Louis and C. Lombart, Journal of Product and Brand Management 19, 114(2010).

9. E. Delgado-Ballester, and J. L. Munuera-Aleman, Journal of Product andBrand Management 14, 187 (2005).

10. M. Horppu, O. Kuivalainen, A. Tarkianen, and H. K. Ellonen, Journal of Prod-uct and Brand Management 17, 403 (2008).

11. A. Mukherjee and P. Nath, European Journal of Marketing 41, 1173 (2007).12. J. J. Zboja and C. M. Voorhees, Journal of Services Marketing 20, 381 (2006).13. H. Y. Ha, Journal of Product and Brand Management 13, 329 (2004).14. R. M. Morgan and S. D. Hunt, Journal of Marketing 58, 20 (1994).15. R. E. Spekman, Strategic Supplier Selection: Understanding Umg Term Buyer

Relationships, Business Horizons, (July/August) (1988), pp. 75–81.16. Srinivasan, Information Management and Computer Security 12, 66 (2004).17. S. S. Alam and N. M. Yasin, Journal of Theoretical and Applied Electronic

Commerce Research 5 (2010).18. M. Mohammadian and M. Ghanbar, Switzerland Research Park Journal 103

(2014).19. N. Ozguven, Chinese Business Review 10, 990 (2011).20. J. C. Roca, J. J. Garcia, and J. J. Vega, Information Management and Com-

puter Security 17, 96 (2009).21. W. D. Salisbury, R. A. Pearson, A. W. Pearson, and D. W. Miller, Industrial

Management and Data Systems 101, 165 (2001).22. C. Flavian and M. Guinaliu, Industrial Management and Data Systems

106, 601 (2006).23. H. H. Chang and S. W. Chen, Online Information Review 32, 818 (2008).24. W. Huang, H. Schrank, and A. J. Dubinsky, Journal of Consumer Behaviour

4, 40 (2004).25. M. Laroche, M. V. Nepomuceno, and R. Marie-Odile, Journal of Consumer

Marketing 27, 197 (2010).26. K. H. Hahn and J. Kim, International Journal of Retail and Distribution

Management 37, 126 (2009).27. L. V. Casalo, C. Flavian, and M. Guinaliu, The International Journal of Bank

Marketing 26, 399 (2008).28. H. Kim and J. Song, Journal of Research in Interactive Marketing 4, 376

(2010).29. C. Kohli, L. Leuthesser, and S. Rajneesh, Got slogan? Guidelines for creating

effective slogans, Business Horizons (2007), pp. 415–422.30. L. C. Harris and M. H. Goode, Journal of Services Marketing 24, 230 (2010).31. Y. H. Chen and S. Barnes, Industrial Management and Data Systems 107, 21

(2007).32. J. F. Hair, Jr., B. Babin, H. A. Money, and S. Philip, Essentials of Business

Research Methods, Wiley: Malloy Inc., USA (2003).


3399




Vol. 21, 3400–3404, 2015

Understanding Attitude of Online Shoppers:

Integrating Technology and Trust Factors

Li Yuan Hui1, Mohd Shoki Md Ariff1�∗, Norhayati Zakuan1, Norzaidahwati Zaidin1,Khalid Ismail2, and Nawawi Ishak3



The online shopping development in China has attracted many researchers to examine technology and humanbehavior factors affecting consumer online attitude in e-commerce setting. In this study, the conceptualizationof technology and trust factors to determine attitude of online shoppers was based on Technology AcceptanceModel (TAM) and Theory of Planned Behavior (TPB). Perceived Ease of Use (PE) and Perceived Useful-ness (PU) of technology factor and trust in website and perceived risk were constructed to determine attitude ofonline shoppers of the biggest network for online shopping in China. A web-based questionnaire was employedto collect 260 feedbacks from online shoppers of the website. Both constructs of technology factor were foundto have significant and positive effect on attitude of online shoppers. The online shopper’s attitude was sig-nificantly and positively influenced by their trust in the website. The effect of perceived risk of trust factor onconsumer attitude was insignificant. This study revealed the importance of integrating technology and trust fac-tors in understanding attitude of online shoppers. This finding calls for online retailers and e-marketers to focuson both technology and human dimensions in their marketing effort to facilitate more people to shop online.

Keywords: TAM, Theory of Planned Behavior, Attitude, Trust, Online Purchase Intention.

1. INTRODUCTIONNowadays, online shopping has become one of the essential

characteristics in the Internet era. Online shopping has become

the third most popular Internet activity after e-mail and web

browsing,1 and it is even more popular than seeking out enter-

tainment, information and news. For online consumers, online

shopping brings a great number and wide range of merchandises

whilst it offers a huge market and numerous business opportu-

nities for online retailers. Online shopping in China is devel-

oping tremendously, in which the online shopping had risen to

142 million, and the volume of the online transaction rose to

RMB523.1 billion in 2010.2 As the world’s largest online pop-

ulation, China is set to become the world’s largest online retail-

ing market.3 Further, the sales volumes online retail market in

China, based on the forecasts by The Boston Consulting Group,

is expected to reach 364 billion US dollars by 2015.4

The online shopping development in China has attracted many

researchers to examine technology and human behavior fac-

tors affecting consumer attitude. Online shopping involves new

technologies in order for online shoppers to browse, search,

compare, and finally making a purchase decision, thus under-

standing how consumers’ perceived the websites technology


and human behavior in online environment are important to

researchers and online retailers.4 In understanding attitude of

online shoppers, the trust and technology factors has been well

studied in on-line shopping and showed that understanding both

the technology and trust factors is important in determining con-

sumer behavioral intention.5 However, most previous studies in

China assessing trust and technology factors and their influence

on consumer attitude addressed the issues separately. This fact

poses some difficulties for e-marketers and online retailers to

incorporate issue of technology and human behavior or trust fac-

tor in designing marketing strategy and program. Thus, integrat-

ing both technology and trust factors is needed to better explain

consumers attitude in China’s online shopping context.

Taoboa6 is one of the shopping website in China with 76.5%

market share in 2009, making it the most preferences website

among online shoppers.2 Since its launching in 2003, Taoboa is

the largest Internet retail and trading website in China, offer-

ing consumers with wide range of products and services.4 It is

one of the world’s largest electronic marketplaces, with over

370 million registered users and annual transaction volume of

almost $15 billion in 2008, equaling approximately one percent

of China’s total retail trade.4 Thus, revealing technology and trust

factors in understanding attitude of online shoppers at the Taoboa



website is important. Therefore, this study contributes to under-

standing predictors of consumer online attitude from both the

human side and technology perspective. For e-marketers and web

designers, this study is significant to them as they have to con-

sider both perspectives in designing the online shopping websites

and promoting them to the consumers.

2. LITERATURE REVIEWOnline shopping research, particularly related to the acceptance

of online shopping technology or system, has been a central

study in e-commerce since 1980s. In the study of acceptance of

online shopping system, it was recognized that two perspectives

have been extensively researches, i.e., the technology and human

behavior. From the technology perspectives, TAM is extensively

used and it is regarded as the most influential model in pre-

dicting consumers’ acceptance of online shopping.7 The technol-

ogy perspective views technology factors, such as PU and PE,8�9

influence attitude of online shoppers. From the human behavior

perspective, TPB10 is mostly used as it provides better under-

standing on how individual attitude could be influenced by trust

factor. However, most researches addressing these two perspec-

tives separately, thus contributing to the technology or human

bias to understanding attitude of online shoppers. For examples,

PU and PE of technology factor were used by Shoki et al.,11 and

Ramayah,12 and Wang and Tseng13 and Zheng et al.14 employed

trust in website and perceived risk of trust factor in examining

consumer online shopping. Van der Heijden et al.,9 examined

online shopping using all those factors and suggest future studies

should be carried out using both perspectives to better under-

standing consumer attitude and behavioral intention in online

shopping scenario. Thus, there is a need to examine consumer

online attitude and purchase intention by combining these two

factors, particularly in the China online shopping market.

In the study of technology factor influencing individual’s

acceptance of information system and technology, such as online

shopping, TAM is one of the most widely used and influential

model.7 TAM is developed by Davis et al.,15 and its goal is to

predict information system acceptance and diagnose problems

related to the users experience with the system. TAM specifi-

cally measures the determinants of computer usage in terms of

PU and PE. According to TAM, a user’s acceptance of an infor-

mation system is dependent on two factors: perceived useful-

ness and perceived ease of use. Together, these factors determine

users’ attitude toward using a technology or an information sys-

tem, such as online shopping website.

Within human behavior perspective, consumer trust has a strik-

ing influence on consumer willingness and attitude to engage in

online business7 because it is vital to the acceptance and con-

tinuous usage of online shopping. TPB serves as a foundation

between belief and trust, attitude and purchase intention of this

study. It has been proven successful in predicting and explaining

human behavior across various information technologies, such as

online shopping.5

Thus, the research framework addressing the technology and

trust perspectives in understanding consumers’ attitude towards

online shopping is established. It is based on TAM and TPB

theory. PU and PE of technology factor and trust in website

and perceived risk of trust factor were identified as predictors to

the attitude of the purchase intention in online shopping setting.

The following section discusses all the variables of the frame-

work and the interrelationship between the constructs.

PU refers to the degree an individual belief that using a

particular technology or system would enhance his or her

performance.8 It is about consumers’ perceptions regarding

the outcome of an experience with an information system or

technology.16 This includes increase efficiency and accuracy in

performing a certain job. In the context of online shopping, online

shoppers have to access a firm’s website to search and purchase a

product. Therefore, PU is expressed as how useful the online sys-

tem or the website technology in performing online purchasing.

The online shoppers PU of the system of online shopping include

quicker, easier and faster purchasing process as compared to that

of conventional or other channels of purchasing.9 Since online

shopping use internet and the website for online transaction, per-

ceived usability of this task is important and it has a significant

effect on e-shop success.2�7 Measure of PU in online shopping

include usefulness of buying the product or service from the

website, quick and fast processing of online transaction.8�9 In

online shopping setting, PU has been found to have positive

effect on consumer attitude towards online shopping.5�9 There-

fore, H1 is proposed: PU positively affect attitude of online shop-

pers towards online shopping.

PE is defined as the “the degree to which an individual; believes

that using a particular system would be free from physical and

mental effort.”8�15 If consumers perceived it is easier to use a

technology or a system to perform a task than other methods,

then they are more likely to accept the system. PE is an impor-

tant construct of online shopping and understanding it effects on

consumers’ attitude and intention toward online shopping is cru-

cial for online sellers. An on line shopping system perceived as

being helpful in facilitating online transaction and easy to use are

typically evaluated more highly and often deemed desirable.17

According to TAM, “ease of use” is particularly of influ-

ence in the early stages of user experience with a technology

or system.8 Following this, Venkatesh18 stated: “With increas-

ing direct experience with the target system, individuals adjust

their system-specific ease of use to reflect their interaction with

the system.” In the online shopping context, when consumers

get more experienced with companies’ websites, they will adjust

their perceptions regarding the “ease of use” of the websites. As

a consequence, they will form a positive attitude in engaging in

the online shopping. Measure of PE for online shopping include

easy use of the website, easy to do what individuals wanted to do

during accessing to the website, clear and understandable of the

site, flexible to interact with, easy to become skillful in using the

website, and easy to learn how to operate the website.7–9 In online

shopping setting, PE has been found to have positive effect on

consumer attitude towards online shopping.7�9 Therefore, H2 is

established: PE positively affect attitude towards online shopping.

Trust plays a key role in creating satisfied and expected out-

comes in e-business transaction. In online purchasing, online con-

sumer trust is crucial because if online shoppers do not trust the

online system or website, they would not commit in online pur-

chase. Gefen et al.19 assert that the present of trust will increase

the consumers’ belief that the e-retailers will not engage in

opportunistic behavior. Egger20 argues that sufficient trust needs

to exist when placing an order online and when the customer

submit his or her financial information and other personal data in

undertaking financial transactions. Trust beliefs positively influ-

ence customer attitude and online purchase intention, in which

3401


the higher the degrees of consumers’ trust the higher will be the

degree of consumers’ purchase intentions.9 In online setting, trust

in website and consumers’ perceived risk is important determi-

nant of consumer attitude towards the online shopping, and it has

positive effect on attitude to use online shopping.5

Trust in website refers to trust in a virtual environment. Trust

exists in many forms, across multiple domains, and at a variety of

levels.21 Measure of trust in website include ability of the website

to act in online shoppers’ best interest, trustful in dealing with

consumers, keeping commitment to user request, capable and

proficient in conducting the online business, willingness of con-

sumers to give their private and credit card information.21 Trust

in website has been found to have a positive effect on attitude

towards online purchasing.5�9�22 Therefore, H3 is proposed: Trust

in website positively affect attitude towards online shopping.

According to Hofstede and Bond,23 the term risk aversion

is defined as “the level to which people are afraid because of

or feel threatened by an ambiguous situation, and have created

beliefs and institutions that try to avoid these.” Lee and Tan24

stated that consumers perceive a higher level risk when purchas-

ing though Internet compared with traditional retail formats. In

marketing, risk perceptions directly affect purchasing and pur-

chase intention.25–27 Therefore, it is the responsibility of online

retailers to eliminate and lower down the risk encountered by

consumers.

Measure of perceived risk of online shopping include amount

of risk encountered, potential loss or gain, positive or nega-

tive situation of buying online and likelihood of making a good

bargain.21 Negative effects from the perception of risk have also

been found to have a negative impact on shoppers’ attitudes

towards online shopping.9�28�29 Thus, it is hypothesize (H4) that

consumers’ perceived risk of online shopping will have negative

effect on their attitude towards online shopping.

There is a debate about precise definitions of attitude. An atti-

tude can be defined as a positive or negative evaluation of people,

objects, event, activities, ideas, or just about anything in your

environment.27�30 This definition of attitude allows for one’s eval-

uation of an attitude object to vary from extremely negative to

extremely positive, but also admits that people can also be con-

flicted or ambivalent toward an object. It shows that an individual

might at different times express both positive and negative atti-

tude toward the same object. Attitude is related to the emotion

and in online purchase, it is referred as consumer’s positive or

negative feelings of the purchasing behavior in internet.31�32

3. METHODOLOGYThis is a descriptive survey research attempting to examine the

combined effect of technology and trust factors on attitude of

online shoppers. In this study, attitude of online shoppers was

examined based on their experience in the usage of Taoboa web-

site. A survey questionnaire was developed to gather primary

data related to the technology and trust factors, the independent

variables to the dependent variable of attitude towards online pur-

chase. Six items of trust in online website and four questions

of perceived risk of trust factor were constructed based on the

study of McKnight, Choudhury, and Kacmar.9�21�27 Four items of

PU and six questions of PE of technology factor were adapted

from Davis8 and the study of Heijden et al.9 and Shoki et al.7

In addition, three items for attitude were adapted from Heijden

et al.9 and Shoki et al.27 All the questions for PU, PE, trust

in website, and attitude were measured using five point Likert

scales of 1 = strongly agree, 2 = Disagree, 3 = somewhat agree,

4 = Agree, and 5 = strongly agree. For perceived risk, all the

four items were measured from 1 to 5 indicating ‘a very big

risk—a very small risk,’ ‘high potential for loss—high potential

for gain,’ ‘a very negative situation—a very positive situation,’

and ‘very unlikely—very likely.’

The population of this research is all online shoppers of

Taobao’s website. The multivariate procedure was applied to

determine the sample size since the total online shoppers of

this website is unknown. The number of total questionsin the

questionnaire was 26, thus a minimum of 260 questionnaires are

needed for 5% margin error and 130 questionnaires are needed

for 10% margin error Hair et al.33 For this study, the web-based

survey was employed by creating the questionnaires and placing

them using https://docs.google.com/forms/d/1ofofzqeMr9x3n0

DmafSDQ0zH11MSZ61XIg5FcOwx1k4/viewform to collect

260 feedbacks from the respondents for 5% margin error. The

link of this questionnaire was shared using ‘we what’ and ‘qq’

because currently Chinese people like using them to communi-

cate to each other. The convenience sampling technique was used

in this research because this method allows the selection of vol-

unteering respondents by depending on their availability and will-

ingness to help in filling the questionnaire. Only respondents who

have experience shopping at this website can proceed with the

whole questionnaire. The data collected was kept automatically

in Google spreadsheet and finally exported to SPSS for further

analysis.

4. RESULT AND ANALYSISPrior to examining the validity of the questionnaire, Kaiser-

Meyer-Olkin (KMO) was performed to ensure adequacy of sam-

ple. The KMO value for the technology and trust factors and

attitude are 0.958 and 0.896, 0.762 and 0.754 respectively. Since

all KMO values were above 0.5 supported by Bartlett’s test of

Sphericity of 0.00, the researcher was allowed to proceed with

Exploratory Factor Analysis (EFA).

In the EFA, components with eigenvalue greater than 1.0 and

factor loading for items that are equal or greater than 0.50 were

retained in this study. Two components, that is PU and PE of

technology factor has been extracted with the cumulative total

variance explains percentage of 80.279%. For trust factor, two

components were extracted, that is, trust in the website and

perceived risk with the cumulative percentage of 80.904%. All

items proposed for these factors were retained since they were

well loaded as in the proposed framework with factor loadings

of more than 0.5. For attitude, only one component has been

extracted with the cumulative percentage of 88.461%. All the

three items for attitude was retained as the factor loading for the

items were >0.5.

The results of normality test, linearity test and multicollinear-

ity test confirmed that all variables used in this study are inde-

pendence free and met the underlying assumptions of correlation

and regression analyses. Further, the reliability test for each con-

struct of the research framework was performed. In this study,

the Cronbach’s alpha values for PU and PE are 0.915 and 0.914,

0.949 and 0.912 for trust in website and perceived risk, and 0.909

for attitude. Thus, construct of all the variables used in this study

are deemed reliable and valid.

3402


To examine how the technology and trust factors affecting con-

sumer online attitude, the multiple regressions was performed.

This research does not consist of unobservable latent variable

and not too complex that requires structural equation modelling

analysis. As shown in Table I, PU (� .198, t2.745, p < .05) and

PE (� .448, t6.149, p > .01) of technology factors positively and

significantly affected consumer online attitude. Further, trust in

website (� .300, t7.119 p < .01) also positively and significantly

affected attitude of Taoboa’s online shoppers. However, the effect

of perceived risk on attitude of online shoppers was insignifi-

cant (�−0�015� t−0�373�p > 0�05). Thus, H1, H2 and H3 are

accepted and H4 is rejected.

5. DISCUSSIONSThis study addresses technology and trust factors in understand-

ing attitude of online shoppers of Toaboa website in China. The

findings of this research provides several important implication

in the study of consumer online attitude from both technology

and trust perspectives. The implications are:

• The construct of technology and trust factors used in this study

are able to predict attitude of online shoppers towards the online

purchase. As discussed in the para 3, the total variance explained

of EFA for technology factor is 80.279% and 80.904% for trust

factor. Thus, the proposed construct of technology and trust fac-

tors provide empirical evidence in the use of both factors to better

understanding online shopping market in China.

• In this research, PU and PE of technology factor and trust

in the website of trust factor were found to have positive and

significant effect on attitude of online shoppers. Specifically, the

positive effect of PU and PE is sync with the previous studies of

Heijden et al.9 The finding of positive effect of trust in website

is consistent with the research of Li et al.5

• This finding calls for online retailers and e-marketers to focus

on both technology and human dimensions in their marketing

effort to facilitate online shopping. The technology perspective

calls for e-marketers to highlight on the ease of use and use-

fulness of online shopping through their websites. From human

behavior perspective, incorporating element of trust would eas-

ily attract online shoppers to shop at the websites because when

they trust the sites, they tend to have positive attitude towards

the online shopping.

• Accepting the H1 and H2 of this study means that when con-

sumers’ perceived online shopping is useful and easy to handle,

then they will form a favorable attitude towards buying products

electronically. Thus, it can be concluded that in order to ensure

online shoppers to shop at the online retailer websites:

(i) the website must not be too complex to operate and if pos-

sible involve just a few clicks in order to complete the online

Table I. Results on the effect of technology and trust factors on atti-

tude of online shoppers.


Constant �727 .212 3�434 �001PU �182 .066 �198∗ 2�745 �006 4.300PE �417 .068 �448∗∗ 6�149 �000 4.386Trust in website �259 .036 �300∗∗ 7�119 �000 1.469Perceived risk −�013 .036 −�015 −0�37 �709 1.264F 142.687 R2 0.691

Note: ∗p < 0�05; ∗∗p < 0�01.

transaction. For website designers, this fact means designing

consumer-friendly website is crucial;

(ii) in in-store purchase, consumers will visit the outlets,

search, check, touch and finally commit in a purchase of a

product. This process may take minutes to hours depending

on type of product purchased. In online buying, if the online

process take less time compared to that of in in-store pur-

chase, consumers may perceived the online purchase to be

faster, time saving, quick and very useful. This point must be

well addressed in marketing promotional mix to attract online

buyers continuing the online purchase.

• The negative effect of perceived risk on consumer attitude

towards online shopping is surprisingly insignificant. This finding

is inconsistent with the previous researches of Van der Heijden

et al.,9 O’Cass28 and Shih.29 This study also addressed trust in

website and this part of trust was positively influenced the atti-

tude of online shoppers. Trust includes consumers’ belief that

Taoboa website is capable and proficient in its business. In con-

trast, some aspects of perceived risk addressed potential loss and

gain and the likelihood of making a good bargain through online

purchase, which is difficult to assess by respondents, thus may

contribute to the insignificant effect of perceived risk on the

attitude.

6. CONCLUSIONSThe issue of how technology and trust factors affect attitude of

online shoppers provide additional insight in the study of attitude

of online shoppers, particularly in China online shopping. These

two factors are proven to have significant effect on the attitude,

and the variation in the attitude of online shoppers is very much

influenced by these two factors. Perceived risk of trust factor is

insignificant, and this requires researchers to further validate this

finding. Sampling procedure pose another limitation of this study

because background of the respondents, particularly frequency

and experience of online shoppers are not considered in the sam-

pling frame and strategy. Previous studies highlighted that these

two aspects influenced perception of online shoppers. Three vari-

ables in this study are related to how consumer perceived ease

of use and usefulness as well as perceived risk, therefore, the

inclusion of these two points would be interesting to research.

References and Notes1. N. Li and P. Zhang, Consumer online shopping attitude and behaviour, Eight

American Conference on Information System (2002), pp. 508–517.2. G. Jun and N. Jaafar, International Journal of Business and Social Science

2, 122 (2011).3. K. Michael, China’s Online Retail Market to Triple to $364bB by 2015, Says

Research Firm, IDG News, April 12 (2012).4. D. Zhao, The Chinese consumer shopping behavior on taobao, unpublished

thesis, Uppsala Universitet (2012).5. T. L. Wang and Y. F. Tseng, International Journal of Digital Society 2, 433

(2011).6. www.taoboa.com.7. M. S. Md Ariff, N. Zakuan, Y. S. Min, K. A. Rahim, and K. Ismail, International

Journal of Information Processing and Management (IJIPM) 4, 48 (2013).8. F. D. Davis, MIS Quarterly 13, 319 (1989).9. H. van der Heijden, T. Verhagen, and M. Creemers, European Journal of IS

12, 41 (2003).10. I. Ajzen, Organisational Behaviour and Human Decision Processes 50, 179

(1991).11. M. S. Ariff, S. M. Yeow, and N. Zakuan, Adv. Sci. Lett. 20, 268 (2013).12. T. Ramayah and J. Ignatius, Impact of perceived usefulness, perceived ease

of use and perceived enjoyment on intention to shop online, 1 (2005).

3403


13. T.-L. Wang and Y. F. Tseng, International Journal of Digital Society (IJDS)2, 433 (2011).

14. L. Zheng, M. Favier, P. Huang, and F. Coat, Journal of Electronic CommerceResearch 13, 255 (2012).

15. F. D. Davis, R. P. Bagozzi, et al., Management Science 35, 983 (1989).16. F. D. Davis, Bagozzi, and Warshaw, Journal of Applied Social Psychology

22, 1111 (1992).17. M. S. Feathermana and P. A. Pavloub, Int. J. Human-Computer Studies

59, 451 (2003).18. V. Venkatesh and F. D. Davis, Management Science 46, 186 (2000).19. D. Gefen, E. Karahanna, and D. Straub, 27, 51 MIS Q (2003).20. K. C. Ling, L. T. Chai, and T. H. Piew, International Business Research 3, 63

(2010).21. McKnight, Choudhury, and Kacmar, Information Systems Research 13, 334

(2002).22. S. L. Jarvenpaa and N. Tractinsky, Information Technology and Management

1, 45 (2000).

23. G. Hofstede and M. H. Bond, Journal of Cross-Cultural Psychology 15, 417(1984).

24. K. S. Lee and S. J. Tan, Journal of Business Research 56, 877 (2003).25. M. S. Md Arif, M. Sylvester, K. A. Rahim, and N. Zakuan, Adv. Sci. Lett.

20, 2319 (2014).26. V. W. Mitchell, European Journal of Marketing 33, 163 (1999).27. M. S. Md Arif, M. Sylvester, N. Zakuan, K. Ismail, and K. M. Ali, IOP

Conference Series: Materials Science and Engineering 58, 1 (2014).28. A. O’Cass and T. Fenech, Journal of Retailing and Consumer Services 10,

81 (2003).29. H. P. Shih, Information and Mgt. 41, 351 (2004).30. P. G. Zimbardo and J. N. Boyd, J. Personality Soc. Psychol. 77, 1271 (1999).31. Y. B. Chiu, C. P. Lin, and L. L. Tang, International Journal of Service Industry

Management 16 (2005).32. I. Ajzen, Annual Review of Psychology 52, 27 (2001).33. J. F. Hair, R. E. Anderson, et al., Multivariate Data Analysis. Prentice-Hall

(1998).


3404




Vol. 21, 3410–3417, 2015

A Benchmark Feature Selection Framework for

Non Communicable Disease Prediction Model

Daniel Hartono Sutanto∗ and Mohd. Khanapi Abd. Ghani

Biomedical Computing and Engineering Technologies (BIOCORE) Applied Research Group;Faculty of Information and Communication Technology, Universiti Teknikal Malaysia Melaka, Melaka, 76100, Malaysia

Non-communicable disease (NCDs) is the most epidemic disease and high mortality rate in worldwide likelydiabetes mellitus, cardiovascular diseases and cancers. NCDs prediction model have problems such as redun-dant data, missing data, imbalance dataset and irrelevant attribute. In data mining, feature selection can handleirrelevant attribute. This paper considers finding the optimal feature selection for NCDs prediction model. Wecomprise 18 feature selection, 4 classification algorithms (Naïve Bayes, Support Vector Machine, Neural Net-work, and Decision Tree) and used 6 NCDs datasets. The result shows that optimally performed feature selectionfor NCDs prediction are weight by SVM, W-Uncertainty, W-Chi, and CBWA.

Keywords: Feature Selection, Prediction, Classification, Benchmark, Comparison, Non-CommunicableDisease, NCDs, Chronic Disease Friedman Test, Nemenyi Test.

1. INTRODUCTIONNon Communicable Diseases (NCDs) or chronic disease are

leading cases of destruction and disability worldwide. NCDs

also known as chronic diseases are a long-lasting condition that

can be controlled, but not immediately cured. Top three main

types of NCDs are diabetes mellitus, cardiovascular diseases and

cancers.1 There is aspect affects the quality of health care, such

as inequity of diagnosis of NCDs due to discrepancy numbers

between patients and doctors.2�3

Guyon has been reviewed feature selection.4 In data mining and

machine learning, feature selection has been utilized for research

and development.5 It has proven in both theory and practice

effectively in increasing predictive accuracy, reducing complex-

ity of learned results, and enhancing learning efficiency. Methods

of individual evaluation rank features according to their impor-

tance in differentiating instances of different classes and able only

remove irrelevant features as redundant features likely have simi-

lar rankings. To set a minimal subset of features using methods of

subset evaluation search, which satisfies some goodness measure

and can remove irrelevant features as well as redundant ones. The

restriction of current research clearly recommends that a differ-

ent framework of feature selection that allows efficient analysis

of both feature relevance and redundancy for NCDs datasets.

2. RELATED WORKSUzer ensemble SFS, SBS and PCA method in feature selection

and used ANN as a classifier.6 Synthetic data have been reviewed


with feature selection methods, and Canedo recommended to use

RelieF.5 To raise the classification accuracy of prediction, we

cause to give the optimal feature selection method in a particular

area. Meanwhile, on that point is no benchmark performance to

compare 18 feature selection techniques for Non Communicable

Disease prediction. This research helps practitioners to remain

abreast of technological advancements in Non Communicable

Prediction. Hence, the research questions of this study are as fol-

lows: Which the feature selection has the optimal performance

for Non-Communicable Disease prediction?

3. METHODOLOGY3.1. A Benchmark Framework

See Figure 1.

3.2. Feature Selection Technique

This experiment used 18 feature selection techniques, their defi-

nitions are briefly stated below.

(1) Weight by SVM. Weight by SVM has a purpose for keep-

ing the highest weighted features in the normal has been inde-

pendently derived in a somewhat different context in dataset.10

The mind is to consider the feature important if it significantly

influences the width of the margin of the ensuing hyper-plane;

this margin is inversely proportional to �w�, the length of w.

Since w = ∑i aixi for a linear SVM model, one can regard

�w�2as a function of the training vectors x1� � � � � xl where xi =

�xi1� � � � � xid�, and thus evaluate the influence of feature j on �w�2

by looking at absolute values of partial derivatives of �w�2 with

respect to xij . (Of course, this ignores the fact that if the train-

ing vectors change, the values of the multipliers ai would also



Non-Communicable DiseaseDataset

WBBC WDBC BUPA

LC ECG PID

Feature Selection Technique

Model Validation

10 Fold Cross Validation

Model Evaluation

Confusion MatrixArea Under Curve

(AUC)

Model Benchmark

Classification Algorithm

DTNNSVMNB

wSVM wPCA wRelief wIG wChi

wCFS wUncertain tTest CBWA

testsign Welch MRweight SAM

MRMR RCCW FCBF PAM SVM-RFE

Difference Test Post Hoc Test

Friedman Test Nemenyi Test

Fig. 1. Benchmark feature selection framework adopted from Ref. [7].

change. Nevertheless, the approach seems appealing.) For the lin-

ear kernel, it turns out that

∑i

��w�2/�xij � = k�wj �

where the sum is over support vectors and k is a constant inde-

pendent of j . Therefore the features with higher �wj � are more

influential in determining the width of the margin. The same rea-

soning applies when a non-linear kernel is used because �w�2

can yet be expressed using only the training vectors xi and the

kernel function.

(2) Weight by Principal Component Analysis. The Weight by

Principal Component Analysis (PCA) operator generates attribute

weights of the given Example Set using a component produced by

the PCA. PCA is a mathematical process that uses an orthogonal

transformation to convert a lot of observations of possibly corre-

lated attributes into a set of values of interrelated attributes called

principal components. The number of principal components is

less than or equal to the number of the original attributes. The

normalize weights parameter is commonly set to true to spread

the weights between 0 and 1. The attribute weights reflect the

relevance of the attributes with respect to the class attribute. The

higher the weight of an attribute, the more relevant it is seen.

Table I. Dataset detail.

Researcher Abbr. Instance Attribute Class Task

[6], [17], [18] WBBC 699 12 2 Classification[18], [19] WDBC 569 32 2 Classification[20], [21] BUPA 345 6 2 Classification[17]–[19] ECG 132 12 2 Classification[18], [22] LC 228 12 2 Classification[19], [23]–[25] PID 768 8 2 Classification

(3) Weight by ReliefF.9 In the ReliefF algorithm, a good dis-

criminating attribute is defined as the attribute that has the same

attribute values in the same class and different attribute values

in different classes. It uses a nearest neighbor method to calcu-

late relevancy scores for each attribute. It measures the worth of

an attribute by repeatedly sampling an instance and computing

given attribute value based on the nearest instance of the same

and different class.

(4) Weight by Information Gain.10 The Information Gain (IG)

is a measure based on Entropy. This univariate filter provides

an ordered ranking of all the features, and then a threshold is

required. In this work, the threshold will be set up selecting the

features which obtain a positive information gain value. The for-

mula for IG is:

InfoGain �Class�Attribute� = H�Class� � H�Class � Attribute�,(3) where H (Class) is the total entropy of the class, and

H�Class � Attribute� is the conditional entropy of the class given

the attribute.

(5) Weight by Chi-Squared.16 The Chi-Squared (X2) method

evaluates each gene individually by measuring the Chi-square

statistics (X2) with respect to the classes. The X2 value of each

gene is computed by

X2 =k∑i=i

n∑j=1

�Aij −Eij �2

Eij

where k is the number of intervals, n is the number of classes,

Aij is the total number of patterns in the ith interval, jth class,

and Eij is the expected frequency of Aij .

(6) Weight by Correlation Based Feature Selection.11 The Cor-

relation based Feature Selection (CFS) was developed by Hall in

1999, which is a heuristic for choosing a subset of features that

are highly correlated with the course of study, yet interrelated

with each other. There are two types CFS, such as Forward Selec-

tion and Backward Elimination. The touchstones for judging the

merit of a subset of features defined as

Ms =ktcf√

k+k�k−1�tff

where Ms is the merit of feature subset S containing k features,

tcf is the average feature-class correlation, and tff is the average

feature–feature correlation.

Table II. AUC evaluation.

AUC Classification Symbol

0.90–1.00 Excellent0.80–0.90 Good0.70–0.80 Fair0.60–0.70 Poor

<0.60 Failure

3411


Table III. Optimal feature selection on wisconsin diagnostic breast cancer.

NB SVM NN DT

FS technique Acc AUC Acc AUC Acc AUC Acc AUC REL

W-SVM 96.14 0.980 97.00 0.996 96.29 0.989 94.99 0.948 1, 3, 6–9W-PCA 95.70 0.989 96.85 0.995 96.14 0.990 95.27 0.965 1–4, 6–8W-ReliefF 96.56 0.984 96.85 0.995 96.28 0.988 95.57 0.954 1–4, 6, 8W-IG 96.57 0.982 96.71 0.994 95.71 0.990 93.71 0.939 1–3, 5–8W-Chi 96.42 0.980 96.85 0.994 96.00 0.991 95.14 0.953 1–8W-CFS 96.57 0.984 96.99 0.994 96.43 0.990 94.84 0.963 1–3, 6–8W-Uncertain 95.70 0.989 96.71 0.994 96.28 0.990 95.42 0.960 1–8t-Test 95.42 0.986 96.70 0.994 96.42 0.990 94.56 0.948 1–9CBWA 95.28 0.985 96.71 0.994 96.00 0.989 94.86 0.945 1–8Test-sign 95.28 0.987 96.71 0.994 95.71 0.986 94.70 0.943 1–8Welch test 95.42 0.985 97.00 0.995 95.29 0.986 94.86 0.942 1–9MR weighting 95.28 0.986 97.00 0.993 95.42 0.988 95.13 0.947 1–8SAM 95.42 0.986 96.57 0.995 95.85 0.985 94.85 0.945 1–8MRMR FS 95.42 0.987 96.85 0.994 96.42 0.988 94.70 0.944 1–8RCCW 95.28 0.988 96.86 0.994 95.28 0.987 94.85 0.945 1–8FCBF 96.14 0.983 96.71 0.994 95.71 0.989 94.85 0.946 1, 2, 4–7PAM 95.42 0.986 96.85 0.995 95.28 0.986 95.57 0.954 1–9SVM-RFE 96.57 0.983 96.71 0.993 96.42 0.990 95.27 0.953 1–4, 6, 8

(7) Weight by Uncertainty (SU).12 Symmetrical Uncertainty is

chosen as the touchstone for assessing the relevance and redun-

dancy of attributes. But those attributes having symmetrical

uncertainty with respect to its class greater than the threshold are

chosen and are considered to relevant. Redundancy is removed by

retaining only those attributes that do not have any approximate

Markov blanket in the current set of remaining attributes.

The Weight by Uncertainty operator calculates the weight of

attributes with respect to the label attribute by measuring the

symmetrical uncertainty with respect to the class. The higher the

weight of an attribute, the more relevant it is seen. The relevance

is calculated by the following formula:

relevance = 2∗ �P�Class�−P�Class � Attribute��/P�Class�

+P�Attribute�

The SU for an attribute is measured by

SU�Class�Attribute�

= 2∗ InfoGain�Class�Attribute�/�H�Class�+H�Attribute�

Table IV. Optimal feature selection on original diagnostic breast cancer dataset.

NB SVM NN DT


W-SVM 95.96 0.989 97.01 0.992 96.83 0.991 95.08 0.931 4, 8, 23, 24, 25, 30W-PCA 94.55 0.988 91.92 0.974 62.74 0.181 92.97 0.916 1, 2, 13, 22, 25W-ReliefF 94.38 0.981 94.56 0.988 97.18 0.988 94.73 0.937 3, 4, 7, 8, 24, 25, 29W-IG 94.56 0.986 94.90 0.988 95.60 0.988 93.84 0.926 24, 25, 29W-Chi 94.56 0.981 95.08 0.989 96.48 0.990 93.33 0.945 3, 8, 24, 25, 29W-CFS 94.55 0.982 94.55 0.984 95.94 0.990 92.80 0.940 3, 8, 24, 25, 29W-Uncertain 94.38 0.986 94.72 0.988 96.66 0.991 95.26 0.955 3, 8, 24, 25, 29t-Test 92.80 0.980 94.03 0.983 96.14 0.989 92.98 0.918 1–7CBWA 92.79 0.980 94.21 0.986 96.66 0.994 93.15 0.914 1–7Test-sign 89.80 0.981 94.03 0.986 95.26 0.989 93.32 0.925 1–9Welch test 92.80 0.984 94.02 0.986 96.13 0.994 88.76 0.875 1–7MR weighting 92.79 0.982 94.19 0.987 95.61 0.992 93.49 0.920 1–8SAM 92.62 0.981 94.03 0.987 95.96 0.992 88.74 0.874 1–7MRMR FS 92.98 0.981 93.86 0.985 95.96 0.993 92.61 0.913 1–8RCCW 92.63 0.984 96.31 0.994 94.19 0.987 92.27 0.903 3, 6, 8, 13, 19, 23, 24, 28, 29FCBF 91.73 0.975 62.74 0.460 62.74 0.077 92.43 0.931 22PAM 92.62 0.981 94.36 0.985 96.49 0.991 88.76 0.876 1–7SVM-RFE 94.55 0.983 94.21 0.988 96.49 0.989 95.43 0.956 7, 8, 24, 25, 29

where H(Class) is the total entropy of the class, and H (Attribute)

is the entropy of the attribute. The SU method is used to com-

pensate for information gain’s bias towards features with more

values.

(8) Ensemble t-test. This operator computes for each attribute a

p-value for 2-sided, 2-sample t-Test. Assumes subpopulation vari-

ances are equal. Degrees of freedom are estimated from the data.

(9) Correlation Based Weak Associations. Correlation Based

Weak Associations (CBWA) is a weighting method based on the

assumption of neighborhood via correlation analysis. CWBA tri-

als to identify significant features. A correlation/similarity matrix

M for all features is constructed. M is repeatedly multiplied by

itself to emphasize examples with a strong membership. At last

M is multiplied with the importance vector displaying the corre-

lation/similarity of each with the label.

(10) Test Significant. This technique computes checks for each

attribute whether class variance is significantly different via

F -test and then choose between t-Test and Welch-test to compute

a p-value.

3412


Table V. Optimal feature selection on BUPA liver disorder dataset.

NB SVM NN DT


W-SVM 56.23 0.578 69.34 0.698 71.60 0.746 61.71 0.511 2–5W-PCA 56.58 0.602 67.82 0.696 73.01 0.752 60.59 0.500 3–5W-ReliefF 58.26 0.578 69.81 0.707 69.64 0.741 59.43 0.500 5W-IG 58.50 0.623 57.97 0.544 71.27 0.777 60.89 0.500 3W-Chi 59.74 0.653 57.97 0.513 70.13 0.763 62.91 0.512 1–6W-CFS 55.98 0.636 68.94 0.694 69.27 0.766 61.17 0.499 1W-Uncertain 60.55 0.639 57.97 0.469 71.02 0.748 62.06 0.509 2–5t-Test 60.60 0.591 57.97 0.433 57.10 0.557 58.55 0.503 1CBWA 57.97 0.581 57.97 0.527 69.61 0.761 62.31 0.516 1Test-sign 58.01 0.598 57.97 0.492 57.34 0.569 58.55 0.500 1Welch test 58.24 0.587 57.97 0.432 59.41 0.561 62.29 0.510 1MR weighting 60.02 0.602 57.97 0.447 56.25 0.558 58.55 0.500 1SAM 59.68 0.600 57.97 0.505 59.14 0.552 58.84 0.500 1MRMR FS 59.43 0.591 57.97 0.494 70.38 0.779 58.55 0.502 1RCCW 60.04 0.600 57.97 0.485 57.06 0.545 62.89 0.510 5FCBF 59.71 0.620 57.97 0.482 59.71 0.636 61.15 0.500 1–6PAM 60.55 0.606 67.54 0.719 69.56 0.758 58.55 0.503 1, 3–6SVM-RFE 60.24 0.621 71.69 0.719 71.61 0.760 62.33 0.509 3

(11) Welch Test. This technique computes for each attribute a

p-value for 2-sided, 2-sample Welch-Test. Does not assume sub-

population variances are equal. Degrees of freedom are estimated

from the data. Another statistical test for measuring significant

differences between the mean of two classes, the Welch-test, is

defined equally

w�x� = x+−x−√∑i+ �xi − x+�

2/n++∑i+ �xi − x+�

2/n−

The Weight by Welch-test-operator1 computes for each feature

a p-value for the two-sided, two-sample Welch-test. It does not

assume subpopulation variances are equal. Degrees of freedom

are estimated from the data.

(12) MR Weighting. This technique calculates weights for all fea-

tures. Selects Pearson correlation, mutual information or F -test

depending on the feature and label type (numerical/nominal).

(13) Significance Analysis for Microarrays. Significance Anal-

ysis for Microarrays (SAM) calculates a weight for numerical

Table VI. Optimal feature selection on echocardiogram dataset.

NB SVM NN DT


W-SVM 94.82 0.993 96.25 1.000 98.57 1.000 97.32 0.500 1, 2, 4, 8, 9W-PCA 97.50 0.987 96.07 0.990 95.89 0.973 96.07 0.500 1, 6W-ReliefF 94.46 0.990 98.57 1.000 98.75 1.000 96.07 0.500 1, 2, 4W-IG 96.07 0.997 94.64 1.000 98.75 0.993 97.50 0.500 1, 2, 8, 9W-Chi 95.89 0.987 97.14 1.000 98.57 1.000 96.07 0.500 1, 2W-CFS 95.89 0.990 93.39 0.950 98.57 1.000 97.50 0.500 1W-Uncertain 94.64 1.000 96.07 0.993 98.75 1.000 97.32 0.500 1, 2, 8, 9t-Test 95.89 0.985 96.07 0.993 98.57 1.000 97.32 0.500 1, 2CBWA 96.07 0.987 97.14 0.990 98.75 0.993 96.07 0.500 1–4Test-sign 97.14 0.990 95.89 0.993 98.75 0.993 97.32 0.500 1–5Welch test 98.57 0.970 97.32 0.990 98.57 1.000 97.50 0.500 1MR weighting 95.71 0.980 96.07 1.000 98.75 1.000 95.89 0.500 1–5SAM 95.89 0.990 96.07 1.000 98.75 1.000 97.14 0.500 1–6MRMR FS 94.64 0.993 95.89 1.000 98.75 1.000 97.50 0.500 1–2RCCW 98.75 0.987 95.89 0.993 98.57 1.000 96.07 0.500 1–2FCBF 97.14 0.980 95.89 0.993 98.57 1.000 95.89 0.500 1–2PAM 97.14 0.990 95.71 1.000 98.75 1.000 97.32 0.500 1–2SVM-RFE 97.32 0.980 96.07 1.000 98.75 0.980 96.07 0.500 1–5

features. For the very high-dimensional problem of analyzing

microarray-data.18 suggests scoring the genes with the SAM

statistic or relative difference d�x� which is defined equally

d�x� = �x+−x−�

×(s0+

√1/nx+1/n−n++n−−2

(∑i+�xi−x+�

2+∑i−�xi−x−�

2

))−1

i+ /− denotes the indices and x+ /− denotes the mean of all

examples belonging to the positive/negative class and s0 is a

small correctional parameter controlling the influence of variabil-

ity. This function is implemented in the Weight by SAM-operator.

(14) MRMR FS.13 Iteratively add the feature with the most infor-

mation regarding the label and the least redundancy to the already

selected features. Implementation of Ding and Peng’s 2003 “min-

imum Redundancy Maximum Relevance Feature Selection.”

3413


Table VII. Optimal feature selection on lung cancer dataset.

NB SVM NN DT


W-SVM 82.41 0.874 87.75 0.941 86.80 0.928 82.06 0.771 1, 3W-PCA 71.48 0.479 72.35 0.512 71.92 0.421 71.92 0.500 9W-ReliefF 81.58 0.873 88.12 0.935 87.67 0.932 81.11 0.732 1, 3, 5W-IG 81.11 0.860 87.75 0.938 87.79 0.938 84.23 0.823 1, 3, 5W-Chi 80.65 0.884 88.18 0.938 86.82 0.927 79.86 0.810 1, 3W-CFS 82.92 0.852 85.55 0.918 86.78 0.920 74.55 0.699 1, 3, 5–8W-Uncertain 82.43 0.874 87.69 0.932 88.62 0.916 82.91 0.788 1, 3t-Test 78.48 0.832 86.86 0.926 88.68 0.942 81.50 0.777 1–3CBWA 78.91 0.826 87.27 0.934 87.67 0.939 75.42 0.729 1–3Test-sign 82.43 0.851 87.27 0.936 88.14 0.924 77.63 0.693 1–3Welch test 82.04 0.847 86.88 0.942 87.67 0.939 79.84 0.804 1–3MR weighting 79.80 0.820 87.79 0.937 87.73 0.932 79.78 0.697 1–4SAM 78.85 0.805 87.73 0.932 88.62 0.928 79.37 0.698 1–4MRMR FS 79.01 0.808 86.84 0.925 87.69 0.932 80.22 0.692 1–3RCCW 81.13 0.848 87.31 0.932 87.25 0.931 80.71 0.786 1–4FCBF 79.35 0.855 86.84 0.938 88.62 0.946 81.13 0.771 1–3PAM 78.95 0.809 87.31 0.929 87.29 0.934 81.56 0.714 1–3SVM-RFE 83.34 0.875 88.14 0.931 88.18 0.926 79.43 0.818 1, 3, 5

(15) Recursive Conditional Correlation Weighting. Recursive

Conditional Correlation Weighting (RCCW) is able to select

top k features with estimated highest conditional correlation to

the label. Performs block wise tournaments between b features.

Removes the feature which has the weakest conditional correla-

tion with the label given the b−1 other attributes. This operation

of block wise elimination is repeated until only k features are

forgotten.

(16) Fast Correlation-Based Filter.14 Fast Correlation-Based Fil-

ter (FCBF) removes irrelevant and redundant features based on

FCBF Algorithm.

(17) Shrunken Centroids/PAM.15 Implicit feature selection and

classification using nearest shrunken centroids. The centroids of

each class is computed and are soft thresholder by the given

shrinkage amount. All those attributes whose centroids after

shrinking, match the overall centroid for each class, do not con-

tribute to the near-centroid computation. The amount of shrink-

age could be chosen by cross-validation, narrowing down to the

value that provides the highest accuracy using the least number

Table VIII. Optimal feature selection on Pima Indian dataset.

NB SVM NN DT


W-SVM 76.30 0.814 77.09 0.827 75.78 0.814 73.05 0.631 1–4, 6W-PCA 74.99 0.801 74.34 0.785 76.30 0.822 72.93 0.632 2, 4, 6W-ReliefF 76.17 0.822 76.45 0.822 75.78 0.803 72.52 0.623 2, 6, 8W-IG 76.69 0.812 76.69 0.819 76.29 0.829 72.92 0.631 2, 6, 8W-Chi 76.82 0.810 77.09 0.818 76.55 0.833 72.65 0.629 2, 6, 8W-CFS 76.70 0.824 76.82 0.817 76.31 0.827 72.53 0.623 2, 6, 8W-Uncertain 76.70 0.827 76.55 0.817 76.29 0.825 72.78 0.634 2, 6, 7, 8t-Test 74.87 0.799 74.34 0.800 75.13 0.806 72.67 0.620 1–3CBWA 74.88 0.797 74.87 0.802 74.48 0.801 72.40 0.624 1–3Test-sign 75.38 0.798 74.61 0.798 75.39 0.803 72.27 0.620 1–3Welch test 75.27 0.792 74.75 0.802 74.74 0.798 73.05 0.630 1–3MR weighting 75.00 0.796 75.27 0.802 74.88 0.804 72.66 0.622 1–3SAM 75.66 0.804 75.01 0.801 74.48 0.802 72.52 0.625 1–2MRMR FS 75.53 0.806 75.01 0.805 75.13 0.801 72.01 0.616 1–2RCCW 75.64 0.806 75.13 0.800 75.78 0.800 72.53 0.626 1–3FCBF 75.39 0.801 76.84 0.821 74.49 0.798 72.53 0.618 1–7PAM 75.01 0.796 75.39 0.804 75.01 0.806 72.27 0.622 1–3SVM-RFE 75.92 0.819 76.70 0.826 75.02 0.804 72.78 0.628 1–4, 6, 8

of attributes. The weight of an attribute is taken to be the num-

ber of classes for which the attribute is relevant, i.e., the number

of classes for which the class centroid of the attribute does not

match its overall centroid.

(18) SVM-RFE.16 Guyon et al. proposed a new gene selection

algorithm of a support vector machine method based on recursive

feature elimination (SVM-RFE).

The SVM-RFE method eliminates iteratively the gene that is less

significant to the classifier from the gene set. The implication of

the factor for the classifier is evaluated by gene ranking score.

The Gene ranking score is determined by the same square of

the weight vector w of the support vector machines, and w is

calculated as

w =n∑

i=1

�iyixi

where xi is the gene expression vector of a sample i in the train-

ing set, yi ∈ �−1�+1� is the class label of sample i, and �i can

3414


Table IX. Summary statistics of 18 feature selection.

No Variable Observations MIN MAX Mean Std. deviation

1 W-SVM 24 0 24 0.500 1.0002 W-PCA 24 0 24 0.181 0.9953 W-ReliefF 24 0 24 0.500 1.0004 W-IG 24 0 24 0.500 1.0005 W-Chi 24 0 24 0.500 1.0006 W-CFS 24 0 24 0.499 1.0007 W-Uncertain 24 0 24 0.469 1.0008 t-Test 24 0 24 0.433 1.0009 CBWA 24 0 24 0.500 0.99410 Test-sign 24 0 24 0.492 0.99411 Welch test 24 0 24 0.432 1.00012 MR weighting 24 0 24 0.447 1.00013 SAM 24 0 24 0.500 1.00014 MRMR FS 24 0 24 0.494 1.00015 RCCW 24 0 24 0.485 1.00016 FCBF 24 0 24 0.077 1.00017 PAM 24 0 24 0.500 1.00018 SVM-RFE 24 0 24 0.500 1.000

be estimated from the training set. Most �i are zero, and only

the training vectors with non-zero �i are support vectors.18

3.3. Classification Algorithm

Based on previous research, the optimally performed classifiers

for non-communicable disease is NB (Naïve Bayes), NN (Neural

Network), SVM (Support Vector Machine), and DT (Decision

Tree).

3.4. Non-Communicable Disease Dataset

NCDs datasets have been picked up from internet reposito-

ries, primarily from the UCI Machine Learning Repository. This

research used 8 secondary datasets which it consists of diabetes,

heart, cancer datasets (Table I).

3.5. Model Validation

This research uses a stratified 10-fold cross-validation for learn-

ing and testing data. This signifies that this research divides the

training data into 10 equal parts and then perform the learn-

ing process 10 times. As presented in Table III, each time this

Table X. Pairwise of Nemenyi post hoc test.

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18

1 0 4�375 2�146 0�813 0�063 2�667 −0�104 4�979 4�292 5�792 4�313 4�625 4�688 4�188 4�042 4�771 3�125 1�4792 −4�375 0 −2�229 −3�563 −4�313 −1�708 −4�479 0�604 −0�083 1�417 −0�063 0�250 0�313 −0�188 −0�333 0�396 −1�250 −2�8963 −2�146 2�229 0 −1�333 −2�083 0�521 −2�250 2�833 2�146 3�646 2�167 2�479 2�542 2�042 1�896 2�625 0�979 −0�6674 −0�813 3�563 1�333 0 −0�750 1�854 −0�917 4�167 3�479 4�979 3�500 3�813 3�875 3�375 3�229 3�958 2�313 0�6675 −0�063 4�313 2�083 0�750 0 2�604 −0�167 4�917 4�229 5�729 4�250 4�563 4�625 4�125 3�979 4�708 3�063 1�4176 −2�667 1�708 −0�521 −1�854 −2�604 0 −2�771 2�313 1�625 3�125 1�646 1�958 2�021 1�521 1�375 2�104 0�458 −1�1887 0�104 4�479 2�250 0�917 0�167 2�771 0 5�083 4�396 5�896 4�417 4�729 4�792 4�292 4�146 4�875 3�229 1�5838 −4�979 −0�604 −2�833 −4�167 −4�917 −2�313 −5�083 0 −0�688 0�813 −0�667 −0�354 −0�292 −0�792 −0�938 −0�208 −1�854 −3�5009 −4�292 0�083 −2�146 −3�479 −4�229 −1�625 −4�396 0�688 0 1�500 0�021 0�333 0�396 −0�104 −0�250 0�479 −1�167 −2�81310 −5�792 −1�417 −3�646 −4�979 −5�729 −3�125 −5�896 −0�813 −1�500 0 −1�479 −1�167 −1�104 −1�604 −1�750 −1�021 −2�667 −4�31311 −4�313 0�063 −2�167 −3�500 −4�250 −1�646 −4�417 0�667 −0�021 1�479 0 0�313 0�375 −0�125 −0�271 0�458 −1�188 −2�83312 −4�625 −0�250 −2�479 −3�813 −4�563 −1�958 −4�729 0�354 −0�333 1�167 −0�313 0 0�063 −0�438 −0�583 0�146 −1�500 −3�14613 −4�688 −0�313 −2�542 −3�875 −4�625 −2�021 −4�792 0�292 −0�396 1�104 −0�375 −0�063 0 −0�500 −0�646 0�083 −1�563 −3�20814 −4�188 0�188 −2�042 −3�375 −4�125 −1�521 −4�292 0�792 0�104 1�604 0�125 0�438 0�500 0 −0�146 0�583 −1�063 −2�70815 −4�042 0�333 −1�896 −3�229 −3�979 −1�375 −4�146 0�938 0�250 1�750 0�271 0�583 0�646 0�146 0 0�729 −0�917 −2�56316 −4�771 −0�396 −2�625 −3�958 −4�708 −2�104 −4�875 0�208 −0�479 1�021 −0�458 −0�146 −0�083 −0�583 −0�729 0 −1�646 −3�29217 −3�125 1�250 −0�979 −2�313 −3�063 −0�458 −3�229 1�854 1�167 2�667 1�188 1�500 1�563 1�063 0�917 1�646 0 −1�64618 −1�479 2�896 0�667 −0�667 −1�417 1�118 −1�583 3�500 2�813 4�313 2�833 3�146 3�208 2�708 2�563 3�292 1�646 0

research took another part of dataset for testing and used the

remaining nine parts for learning. Subsequently, this researchers

calculated the mean values and the deviation values from the ten

different testing solutions. This research employs the stratified

10-fold cross validation, because this method has become the

standard and state-of-the-art validation method in virtual condi-

tions. Some trials have also indicated that the role of stratification

improves results slightly.26

3.6. Model Evaluation

This research applies Area Under Curve (AUC) as an accurate

indicator in our experiments to assess the public presentation of

the classification algorithm. AUC is an area under the ROC curve.

Later on just about research, Lessmann et al.27 and Li et al.20 put

forward the use of the AUC to improve cross study comparison.

The AUC has the benefit to improve convergence across empirical

experiments significantly, because it separates predictive perfor-

mance from operating conditions, and presents a general standard

of predictive. A rough guide for classifying the accuracy of a

diagnostic test using AUC is the traditional system, presented by

Belle.28 In the proposed framework, this research added the sym-

bols for easier reading and understanding of AUC (Table III).

3.7. Model Benchmark

In that respect are three families of statistical tests that can be

used for benchmarking two or more classifiers over multiple

datasets:

(1) Parametric tests (the paired t-test and ANOVA), non-

parametric tests (the Wilcoxon and the Friedman test)

(2) The non-parametric test that assumes no commensurability

of the results (sign test).

In Demsar’s research, he recommends the Friedman test for

multiple benchmark classifiers, which relies on less restrictive

assumptions.29 Grounded on this recommendation, in benchmark

framework Friedman test is used to compare the AUCs in differ-

ent classifiers. The Friedman test is computed along the average

ranked (R) performances of the classification algorithms on each

dataset.

3415


Table XI. P-value of Nemenyi post hoc test.

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18

1 1 0.288 0.996 1.000 1.000 0.961 1.000 0.109 0.322 0.020 0.314 0.199 0.180 0.368 0.436 0.157 0.856 1.0002 0.288 1 0.994 0.675 0.314 1.000 0.249 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 0.9203 0.996 0.994 1 1.000 0.997 1.000 0.993 0.933 0.996 0.634 0.996 0.981 0.975 0.998 0.999 0.966 1.000 1.0004 1.000 0.675 1.000 1 1.000 0.999 1.000 0.378 0.714 0.109 0.705 0.551 0.519 0.761 0.819 0.477 0.991 1.0005 1.000 0.314 0.997 1.000 1 0.969 1.000 0.122 0.349 0.023 0.340 0.220 0.199 0.397 0.467 0.174 0.876 1.0006 0.961 1.000 1.000 0.999 0.969 1 0.945 0.991 1.000 0.856 1.000 0.999 0.998 1.000 1.000 0.997 1.000 1.0007 1.000 0.249 0.993 1.000 1.000 0.945 1 0.090 0.280 0.016 0.272 0.169 0.152 0.322 0.387 0.131 0.819 1.0008 0.109 1.000 0.933 0.378 0.122 0.991 0.090 1 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 0.999 0.7059 0.322 1.000 0.996 0.714 0.349 1.000 0.280 1.000 1 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 0.93710 0.020 1.000 0.634 0.109 0.023 0.856 0.016 1.000 1.000 1 1.000 1.000 1.000 1.000 1.000 1.000 0.961 0.31411 0.314 1.000 0.996 0.705 0.340 1.000 0.272 1.000 1.000 1.000 1 1.000 1.000 1.000 1.000 1.000 1.000 0.93312 0.199 1.000 0.981 0.551 0.220 0.999 0.169 1.000 1.000 1.000 1.000 1 1.000 1.000 1.000 1.000 1.000 0.84913 0.180 1.000 0.975 0.519 0.199 0.998 0.152 1.000 1.000 1.000 1.000 1.000 1 1.000 1.000 1.000 1.000 0.82714 0.368 1.000 0.998 0.761 0.397 1.000 0.322 1.000 1.000 1.000 1.000 1.000 1.000 1 1.000 1.000 1.000 0.95515 0.436 1.000 0.999 0.819 0.467 1.000 0.387 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1 1.000 1.000 0.97316 0.157 1.000 0.966 0.477 0.174 0.997 0.131 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1 1.000 0.79517 0.856 1.000 1.000 0.991 0.876 1.000 0.819 0.999 1.000 0.961 1.000 1.000 1.000 1.000 1.000 1.000 1 1.00018 1.000 0.920 1.000 1.000 1.000 1.000 1.000 0.705 0.937 0.314 0.933 0.849 0.827 0.955 0.973 0.795 1.000 1

Let r ij be the rank of the j-th of C algorithms on the i-th

of D datasets. The Friedman test has an aim to compare the

mean ranks of algorithm Rj = �1/D�∑D

i−1 rij . Under the null-

hypothesis, which says that all the algorithms are equivalent and

so their ranks Rj should be clean. The statistic of Friedman is

calculated as follows, and disseminated according to x2F with C

−1 degrees of freedom, when variable D and C are big enough.

x2F = 12D

C�C+1�

[ D∑j

R2j −

C�C+1�2

4

](23)

If the null-hypothesis is rejected, it can go along with a post-hoc

test. When all classifiers are compared to each other, the Nemenyi

test should be applied. Two classifiers have significantly different

performance if the corresponding average ranks differ by at least

the critical difference, shown by

CD = qa

√C�C+1�

D(24)

where critical values qa are based on the studentized range

statistic.

Table XII. Significant differences of nemenyi post hoc test.

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18

1 N N N N N N N N N Y N N N N N N N N2 N N N N N N N N N N N N N N N N N N3 N N N N N N N N N N N N N N N N N N4 N N N N N N N N N N N N N N N N N N5 N N N N N N N N N Y N N N N N N N N6 N N N N N N N N N N N N N N N N N N7 N N N N N N N N N Y N N N N N N N N8 N N N N N N N N N N N N N N N N N N9 N N N N N N N N N N N N N N N N N N10 Y N N N Y N Y N N N N N N N N N N N11 N N N N N N N N N N N N N N N N N N12 N N N N N N N N N N N N N N N N N N13 N N N N N N N N N N N N N N N N N N14 N N N N N N N N N N N N N N N N N N15 N N N N N N N N N N N N N N N N N N16 N N N N N N N N N N N N N N N N N N17 N N N N N N N N N N N N N N N N N N18 N N N N N N N N N N N N N N N N N N

3.8. Experimental Infrastructure

In this research, the experiment equipped with infrastructure con-

sists RapidMiner Toolkit is an open-source system consisting

of a bit of data mining algorithms to automatically examine a

large data collection and extract useful knowledge. The XLSTAT

statistical analysis add-in offers a wide variety of purposes to

enhance the analytical capacities of Excel, producing it the ideal

puppet for your everyday data analysis and statistics prerequi-

sites. The hardware used CPU: HP Z420 Workstation, Proces-

sor: Intel® Xeon® CPU E5-1603 @ 2.80 GHz, RAM: 8,00 GB,

and OS: Windows 7 Professional 64-bit Service Pack 1.

4. EXPERIMENTAL RESULTSIn this section, the performance results of each algorithm on

each dataset will be discussed and research questions will be

answered accordingly. The research will experiment with 4 NCDs

datasets by using 4 classification algorithms and 18 feature selec-

tion techniques. The significant accuracy and AUC shown at

Tables IV–IX. The optimal performance show with black and

gray highlight.

3416


Friedman’s test used Q (Observed value) 57.921, Q (Critical

value) 27.587, DF 17, p-value (Two-tailed) < 0.0001, and alpha

0.05. It has shown critical difference: 5.3764. P -value results of

Nemenyi post hoc test are shown in Table XI. P -value < 0.05

results are highlighted with boldfaced, hence there is a statisti-

cally significant difference between two classification algorithms,

in a column and a row.

From P -value analysis, there is a significant difference

between 18 feature selection techniques. This affects the accuracy

of NCDs prediction model. Significant difference table resulted

from Nemenyi post hoc test shown in Table X. The signifi-

cant feature selection are W-SVM, W-Uncertainty, W-Chi, and

CBWA, it’s indicated in Table XIII.

5. CONCLUSIONNon-communicable Disease (NCDs) is the most epidemic dis-

ease and high mortality rate in worldwide likely diabetes melli-

tus, cardiovascular diseases, and cancers. NCDs prediction model

have problems such as redundant data, missing data, imbalance

dataset and irrelevant attribute. In data mining, feature selec-

tion can handle irrelevant attribute. This paper considers finding

the optimally feature selection for NCDs prediction model. We

comprise 18 feature selection, 4 classification algorithms (Naïve

Bayes, Support Vector Machine, Neural Network, and Decision

Tree) and used 6 NCDs datasets. The result shows that opti-

mally performed feature selection for NCDs prediction is weight

W-SVM, W-Uncertainty, W-Chi, and CBWA.

Acknowledgment: This work was supported with a

grant from LPDP Minister of Finance of Indonesia No.

Kep56/LPDP/2014.

References and Notes1. WHO, Global Status Report on Noncommunicable Diseases (2010).2. M. K. A. Ghani, R. K. Bali, R. N. G. Naguib, I. M. Marshall, and N. S.

Wickramasinghe, Int. J. Healthc. Technol. Manag. 11, 113 (2010).3. D. H. Sutanto, N. S. Herman, and M. K. A. Ghani, Adv. Sci. Lett. 20, 1740

(2014).4. I. Guyon, J. Mach. Learn. Res. 3, 1157 (2003).5. V. Bolón-Canedo, N. Sánchez-Maroño, and A. Alonso-Betanzos, Knowl. Inf.

Syst. 34, 483 (2013).6. N. Yilmaz, O. Inan, and M. S. Uzer, J. Med. Syst. 38, 48 (2014).7. R. S. Wahono, N. S. Herman, and S. Ahmad, Adv. Sci. Lett. 20, 1945

(2014).8. Y.-W. Chang and C.-J. Lin, Causation Predict. Chall. 53 (2008).9. I. Kononenko, Machine Learning: ECML-94 784, 171 (1994).

10. M. A. Hall and L. A. Smith, Comput. Sci. 98, 181 (1998).11. M. A. Hall, Methodology 21i195-i20, 1 (1999).12. L. Yu and H. Liu, Efficient Feature Selection via Analysis of Relevance and

Redundancy 5, 1205 (2004).13. C. Ding and H. Peng, J. Bioinform. Comput. Biol. 3, 185 (2005).14. L. Yu and H. Liu, Int. Conf. Mach. Learn. 1 (2003).15. R. Tibshirani, T. Hastie, B. Narasimhan, and G. Chu, Proc. Natl. Acad. Sci.

U. S. A. 99, 6567 (2002).16. I. Guyon, J. Weston, and S. Barnhill, Mach. Learn. 46, 389 (2002).17. P. Luukka, Expert Syst. Appl. 38, 4600 (2011).18. S. Belciug and F. Gorunescu, J. Biomed. Inform. 52, 329 (2014).19. L.-Y. Chuang, C.-H. Yang, K.-C. Wu, and C.-H. Yang, Comput. Biol. Med.

41, 228 (2011).20. D.-C. Li, C.-W. Liu, and S. C. Hu, Artif. Intell. Med. 52, 45 (2011).21. Y. J. Fan and W. A. Chaovalitwongse, Ann. Oper. Res. 174, 169 (2010).22. N. A. Mat Isa and W. M. F. W. Mamat, Appl. Soft Comput. 11, 1457 (2011).23. M. A. Chikh, M. Saidi, and N. Settouti, J. Med. Syst. 36, 2721 (2012).24. F. Beloufa and M. A Chikh, Comput. Methods Programs Biomed. 112, 92

(2013).25. J. Zhu, Q. Xie, and K. Zheng, Inf. Sci. (Ny) 292, 1 (2015).26. Data Mining Pratical Machine Learning Tools and Techniques, Ian H. Witten

Eibe Frank Mark A. Hall, 3rd edn. (2011).27. S. Lessmann, B. Baesens, C. Mues, and S. Pietsch, IEEE Trans. Softw. Eng.

34, 485 (2008).28. V. Van Belle and P. Lisboa, Artif. Intell. Med. 60, 53 (2014).29. J. Demšar and J. Demšar, J. Mach. Learn. Res. 7, 1 (2006).


3417




Vol. 21, 3418–3421, 2015

Website Quality and Consumer Attitude of Online

Shopping; The Y-Generation Perspective

Yap Soon Jing1, Norzaidahwati Zaidin1, Mohd Shoki Md. Ariff1�∗, Norhayati Zakuan1,Khalid Ismail2, and Nawawi Ishak3



Online shopping is now a trend in e-business across the world, and the website is the main platform for thee-business. Thus, website quality is one of the crucial factors to determine the success of online business.This study aims to determine the website quality and how it fosters the attitude of online shopping amongyoung generation (Y-Gen). Based on the review of the past researches, WebQual was identified as relevantand the framework of this research was developed based on it. Altogether, 39 items of four website qualitywere treated as the key constructs (Usefulness (U), Ease of Use (EoU), Entertainment (E) and ComplimentaryRelationship (CR)). Stratified sampling method and questionnaire were adopted as the sampling procedure anddata collection purposes. The result showed that the four dimensions of website quality were positively andsignificantly affect attitude of Y-Gen towards the online shopping and EoU appeared to be the most dominantpredictor of attitude of online shoppers. Theoretical and managerial implications of the study are presented inthis article.

Keywords: Website Quality, WebQual, Consumer Attitude, Online Shopping.

1. INTRODUCTIONGeneration cohort is a group of people who has similar experi-

ence and event such as common social, political, historical and

economic environment at a similar age.1 Generation Y (Gen

Y) is a cohort who is the most active users of online and

mobile advanced technology and is more frequent in online

activities such as social networks (86%), podcasts (57%), and

text messaging (96%) than any other generation.2 Wolburg and

Pokrywczynski3 found that the buying behavior of Gen Y was

mainly influenced by the internet as compared to other media.

Within the Gen Y group, students of the higher institutions are

categorized as the heavy users of internet.4 Similarly, Delafrooz

et al.5 believed that students are the active internet users and

potentially proficient in utilizing the internet services, such as

online shopping. In term of discretionary spending, this gen-

eration spent nearly $69 billion annually.6 On the same mark,

Lester et al.7 have identified 91% of the college age market

accomplished online purchases, and a quarter of them spent

over $500 per year in online shopping. In statistic, Gen Y

is amounted approximately 6.2 million or 27% of the total

Malaysian population.8 This indicates that Gen Y is a poten-

tial huge segment for online shopping and they will dominate


the online shopping future market. In line with that notion, a

research has suggested for Malaysia online retailers to target

the Malaysian students’ as they possess positive attitude towards

online shopping.9

Online shopping has change the tradition customer-to-

employee interaction into customer-to-website interaction. Since

the website is the medium that represents company for commu-

nication purposes, therefore, a well-developed website is needed,

to lead to success in the virtual market. This reflects the quality

perceived by user that inspires them to continue visiting the site.

Barnes and Vidgen10 stated that the success of company come

from the intention of customers to comeback, which in this sce-

nario, is the intention to revisit the same website. The underlying

message is that, the success of company relies on the repeat cus-

tomers who are loyal to them. Thus, website quality should be

examined continuously because it plays an important role as the

pull factor for online shopping. The high quality website will

be the main reason for consumers purchase decision, and there-

fore online shopping success is significantly impact by website

quality.11 In fact, quality of website is a dominant factor that

drives online business.12

Previous researchers have found that the interaction between

online users and their e-vendors able to develop positive inten-

tion and favorable attitude towards online shopping.13 Despite of



all the claims above, however, there are still lack of study that

clearly identified the relationship between website quality and the

attitude towards online shopping specifically among the Y Gen.

In this case, how Gen Y perceived website quality have to be

investigated and their attitude toward online shopping should be

examined. Therefore, this study contributes to the

(i) Better understanding of website quality and attitude of online

shoppers based on generation cohort of consumer;

(ii) understanding the suitability of the current construct of web-

site quality from the Gen Y perspective who is regarded as the

most active users of online and mobile advanced technology; and

(iii) understanding of buying behavior of a generation (Gen Y)

who is highly active in internet-based shopping, such as

social networks, podcasts, and text messaging than any other

generations.

2. LITERATURE REVIEWYoo14 (2001) defined internet shopping site as the place that

all transactions and relevant activities take place in internet, via

online sellers’ websites. In online shopping, a good company

website is important because it can attract buyers to shop online.

Zeithmal15 (2002) defined website quality as a website that per-

formed well and efficient in shopping, buying and delivery of

products and services. It is about how online buyers evaluate

the features of a website in meeting their expectations and con-

clude the overall performance of the website.16 Website qual-

ity helps customers in making purchase decisions and, therefore

it will lead the company to achieve the sustainable competitive

advantage.17

A number of instruments were developed to measure website

quality of online sellers from the online buyers’ perspective. For

example, Loiacono et al.18 developed a website quality instru-

ment called WebQual and Yoo and Donthu14 proposed SITE-

QUAL in order to measure the perceived quality of an Internet

shopping site. Barnes and Vidgen10 develop eQual 4.0 based on

the evaluation of students towards online book store whist Bauer

et al.19 had developed eTransQual. Wolfinbarger and Gilly20 has

developed. comQ and eTailQ.

WebQual was established originally based on the Technology

Acceptance Model (TAM); perceived usefulness, ease of use and

enjoyment.21�22 TAM is a further extension of Theory of Rea-

soned Action (TRA) which defined an individuals’ acceptance of

information technology. Davis21 stated that the beliefs about sys-

tem utilization behavior will be achieved due to the innovation as

an instrumental in the development of attitude. Previous studies

utilizing TAM in the information systems area highlighted that

there were a direct positive relationship between website effec-

tiveness and attitude towards online shopping.23

Four key constructs of WebQual are Usefulness, Ease of Use,

Entertainment and Complimentary Relationship, and these four

dimensions are correlated with acquisition and visit intentions.18

Table I provides details explanation of these four constructs of

WebQual.

Perceived usefulness is defined as the degree to which the

performance of an activity can be enhanced by technology.21

If online shopping is perceived to be more efficient in term of

time taken to make a purchase, less effort to place an order and

then received the purchased product, structured communication,

secure communication, then online buyers are willing to commit

in on line buying.

Table I. The WebQual-loiacono, watson and goodhue, 2007.

Initial higherlevel category Dimension Description

Usefulness Informational fit-to-ask The information providedmeets task needs andimproves performance

Tailored communica-tion/interactivity

Structured communicationbetween buyer andseller

Trust Secure communicationand observance ofinformation privacy

Response time Time to get a responseafter a query or arequest

Ease of use Ease of understanding Easy to read andunderstand

Intuitive operation Easy to operate andnavigate

Entertainment Visual appeal The aesthetics of a website

Innovativeness The creativity anduniqueness of sitedesign

Flow emotional appeal Individual emotionalintensity of involvementwith the website

Complementaryrelationship

On-line completeness Provision for all necessarytransaction to becompleted online

Better than alternativechannel

Better option than othermeans of interactingwith the company

Consistent image Compatibility of thewebsite image with theimage of thefirm/product it isadvertising

Perceived ease of use is defined as the concentration of phys-

ical and mental efforts that a user assumes to receive when

considering the use of technology.18 It deals with perception of

consumers about clear and comprehensive website provided by

retailer, less mental effort to use the website, and shop according

to their ways.24

Entertainment of a website is also important factor contribut-

ing to the better assessment of a website quality. Entertainment

involves visual appealing of a website, innovativeness and flow-

emotional appeal.18 It is about the aesthetic value features in a

website, complementing the functionality of the website.

Complimentary relationship involves consistent image of a

company website, completeness of online transaction and use of

online shopping is better than other alternative channels.18 Con-

sistent image involves how the website projects an image con-

sistent and match with the company’s image and the website fits

with users’ expected image of the company. Completeness of

online shopping transaction will influence users’ judgement on

quality or performance of the website.

In examining the relationship between website quality and atti-

tude of online shoppers, perceived ease of use and perceived

usefulness were dominant predictors of attitude towards online

shopping.24 Users’ attitude is directly affected by usefulness and

ease of use of the system.21�25�26 In fact, both usefulness and

ease of use of the system is directly influence the attitude, in

term of how easy and how useful online shopping sites are in

creating a good shopping environment.27 Perceived usefulness is

3419


the main determinant of attitude towards online shopping28 and

it is significantly related to attitude towards online shopping.27

Entertainment and trust are important predictors of perceived

value while online completeness (dimensions under Complimen-

tary Relationship) and entertainment were important predictors

of loyalty intentions.18 Thus, it can be concluded that factors

affecting attitude towards online shopping are not only ease of

use and usefulness, but also enjoyment/entertainment, complete-

ness of online transaction, better interaction with the company

and compatibility of the website image with the image of the

firm/product.18�29

Figure 1 depicted the conceptual model of the study derived

from the above discussion. The model hypothesized the four con-

structs of WebQual–Usefulness, Ease of Use, Entertainment and

Complimentary Relationship–will have positive and significant

effect on Gen Y online shopping attitude.

3. METHODOLOGYThis research employed descriptive survey approach to examine

the website quality and how it affects the attitude towards online

shopping among the Generation Y. Current trend in the usage

of Malaysian-based website and online shopping among Gen Y

have been reviewed in order to examine website quality key

constructs affecting attitude towards online shopping. Review of

theories related the website quality had contributed to the identi-

fication of four key constructs in the WebQual instrument which

are Usefulness, Ease of Use, Entertainment and Complimentary

Relationship (independent variables). The impact of the web-

site quality must be examined to determine the extent to which

they exert influence on attitude of online buyers. Thus, attitudes

towards online shopping were established as dependent variable

of this study.

Based on extensive review on theories and previous researches

related to determinants of website quality, the researcher devel-

oped the questionnaire that consisted of 39 items that were made

up from the four website quality key constructs. These items

were based on the WebQual dimensions and were adapted from

the researches of Loiacono, Watson and Goodhue18 and Corbitt,

Thanasankit and Yi.30 Five items of attitude of online shopping

were based on the work of Han, Tibert and Marcel.31

In this research, all measurements are based on five-point

Likert scales ranging from “strongly disagree” (1) to “strongly

agree” (5).

All 715 business students32 in a faculty of a public university

serve as the population of this study. They are mainly in the Gen

Y group (by age). A total of 248 questionnaires were distributed

according to the determined sample size (according to the table

proposed by Krejcie and Morgan.33 Stratified sampling technique

was adopted, since the number of population is known, and the

Website Quality

Usefulness

Online attitudeof Gen Y

Ease of Use

Entertainment

Complimentary Relationship

Fig. 1. The conceptual model.

sample was determined according to the strata, which are the year

of study: year one, year two, year three and year four in order

to increase the accuracy of the data and the representativeness of

the population in this study.

Prior to multiple regression analysis, researcher performed the

diagnostic tests. The results showed all the assumptions: linearity,

normality and multicollinearity were met. For reliability analysis,

the Cronbach’s Alpha for Ease of use was 0.849, 0.840 (Enter-

tainment), 0.808 (usefulness) and 0.7 (Complimentary relation-

ship). The perceived alpha value for attitude is 0.893. Thus, all

key constructs were statistically reliable and the level of reliable

was good and acceptable.

The Exploratory Factor Analysis performed produced the val-

ues of Kaiser-Meyer-Olkin Measure of Sampling were greater

than 0.6 which were 0.756 (website quality) and 0.725 (Con-

sumer attitude) with Barlett’s test of Sphericity values of 0.000

that were less than 0.05, indicating that the proportion of vari-

ance in the both independent and dependent variables was caused

by underlying factors and able to proceed with factor analysis.

The total Variance Explained of Website Quality is 74.351% with

four factors with eigenvalues more than 1–Usefulness, Ease of

Use, Entertainment and Complimentary Relationship. In Rotated

Component Matrix, all items were of the four dimensions and

they were retained because of factor loading more than 0.5. All

factor loading for the five items of attitude of the Y-Gen were

more than 0.5 and they are retained for further analysis with the

Total Variance Explained of 63.561%.

4. RESULT AND ANALYSISMultiple Regressions is shown in Table II, the result of Useful-

ness, Ease of Use, Entertainment and Complimentary Relation-

ship were significant with the significant value of less than 0.05

when the impact of website quality key constructs towards online

shopping attitude was analyzed. This shows that all the dimen-

sions of website quality have significant affect towards the online

shopping attitude of Gen Y. Ease of use (� 0.320, p < 0�001)

exerted the highest effect on the Gen Y attitude. As a result,

the website quality key constructs were positively influencing the

attitude towards online shopping.

5. DISCUSSIONSThe main finding of this research is that, the website quality

positively and significantly affected attitude of Gen Y. This effect

is also observed in other researches, however among the general

online users.18�21�22 Thus, it can be concluded that regardless of

generation of online buyers, perceived usefulness of a website,

Table II. Multiple regression result between website quality and con-

sumer attitude (dependent variable) of online shopping.


Constant 0�878 0.289 3.036 0.003Usefulness 0�281 0.068 0.197∗∗ 4.146 0.001 1.183Ease of use 0�380 0.064 0.320∗∗ 5.097 0.001 1.532Entertainment 0�219 0.064 0.175∗∗ 3.422 0.001 1.370Complimentary

relationship0�383 0.072 0.283∗∗ 5.297 0.001 1.495

F 18�602 R 0.732 R2 0.536

Note: ∗p < 0�05; ∗∗p < 0�01.

3420


entertainment, complimentary relationship and ease of use are

very crucial in determining attitude of online buyers.

Among the four construct of website quality, ease of use was

found to have stronger effect on attitude of Gen Y in the online

shopping setting. This finding synchronises with what had been

addressed by Yulihasri and Daud28 and a research by Lim and

Ting.27 The underlying factors are that online buyers tend to have

positive attitude towards the online shopping if they feel that the

online shopping website is easy for them to search a product,

place an order and deal with the online sellers. This phenomenon

is universal as it was evidenced in the many previous studies as

mentioned earlier. Thus, regardless of generation of users, how

online buyers perceived easiness of accessing and using the web-

site will contribute to the positive attitude towards the online

shopping. According to WebQual, Ease of Use is measure web-

site the degree of easy to read, operate and understand. The cru-

cial factors of information quality are the understandability and

format of information which is the way information is presented

to consumers.34 Thus, the content of website easy to read and

understand could result consumer perceived that the information

provided by the website to be high quality. This has suggested

that online shopping company could emphasize Ease of Use to

improve the quality of the website.

Childers et al.24 stated perceived usefulness is the stronger

predictors of consumer attitude towards online shopping. This

study supported the idea in which perceived usefulness is the key

construct that help users perform the task efficiently in term of

time and communication, consistent with the findings of Gefen25

and Bisdee.35 It shows that Gen Y will prefer to use the web-

site if the website could enable them to accomplish their task.25

Similar finding is observed among the Gen Y, therefore, it can

be concluded that this phenomenon is universal among online

users.

6. CONCLUSIONSThe result has shown that the website quality positively affect

Gen Y attitude towards online shopping. This study concludes

that it supported major theories addressing website quality and

attitude of online buyers like TAM and TRA. It is suggested that

future study to extend other groups of Gen Y prior to gener-

alizing the universal issue of website quality and online buyers

attitude. Further, this conclusion is made by comparing this find-

ing with previous researches involving general users. Therefore,

comparing among generation of online buyers is recommended.

References and Notes1. K. C. Williams and R. A. Page, Journal of Behavioural Studies in Business 1

(2010).2. Consumer Behavior Report, Online Purchasing Trends by Generation from

http://www. PriceGrabber.com (2008), p. 1.

3. J. M. Wolburg and J. Pokrywczynski, Journal of Advertising Behaviour 4, 33(2001).

4. S. Jones, The Internet goes to college: How students are living in the futurewith today’s technology. Pew Internet and American Life Project, September(2002), Available at: www.pewinternet.org.

5. N. Delafrooz, L. Paim, and A. Khatibi, Journal of American Science 6, 137(2010).

6. E. Wong, College back to school spending up 13%. Brandweek July (2010),Vol. 7.

7. D. Lester, A. M. Forman, and D. Lloyd, Services Marketing Quarterly 27, 123(2006).

8. Economic Planning Unit Malaysia from (2010). http://www.epu.gov.my/html/themes/epu/epu/images/common/pdf/ec Retrieved December 2013,

9. M. F. Sabri, M. MacDonald, J. Masud, L. Paim, T. Hira, and M. Othman,Consumer Interests Annual 54, 166 (2008).

10. S. J. Barns and R. T. Vidgen, Journal of Electronic Commerce Research3, 114 (2002).

11. Li, Hairong, T. Daugherty, and F. Biocca, Journal of Interactive Marketing15, 13 (2001).

12. B. Bai, R. Law, and I. Wen, International Journal of Hospitality Management27, 391 (2008).

13. D. Cyr, Journal of Management Information System 20, 33 (2008).14. Y. Boonghee and N. Donthu, Quarterly Journal of Electronic Commerce 2, 31

(2001).15. V. Zeithaml, Managing Service Quality 12, 135 (2002).16. A. M. Aladwania and P. C. Palvia, Information and Management 39, 467

(2002).17. J. Santos, Management Service Quality 13, 233 (2003).18. E. T. Loiacono, R. T. Watson, and D. L. Goodhue, WebQual: A web

site quality instrument, Working Paper 2000-126-0, University of Georgia(2002).

19. H. H. Bauer, T. Falk, and M. Hammerschmidt, Journal of Business Research59, 866 (2006).

20. M. Wolfinbarger and M. C. Glly, Comq: Dimensionalizing, Measuring and Pre-dicting Quality of the E-tailing Experience, Working Paper No.02–100, Mar-keting Science Institute (2002).

21. F. D. Davis, MIS Quarterly 13, 319 (1989).22. J. C. Lam and A. Parasuraman, Journal of Interactive Marketing 22, 19

(2008).23. J. W. Moon and Y. G. Kim, Extending the TAM for a World-Wide-Web cntext,

Inofrmation and Management (2001), Vol. 38, pp. 217–230.24. T. L. Childers, L. C. Carr, J. Peck, and S. Carson, Journal of Retailing 77, 511

(2001).25. D. Gefen, D. W. Straub, and M. C. Boudreau, Communications of the Associ-

ation for Information System 4, 2 (2000).26. Z. Selamat, N. Jaffar, and B. H. Ong, European Journal of Economics,

Finance and Administrative Sciences 1, 143 (2009).27. W. M. Lim and D. H. Ting, Journal of Modern Applied Science 6, 49

(2012).28. I. A. Yuslihasri and A. K. Daud, International Journal of Marketing Studies

3, 128 (2011).29. T. P. Monsuwe, G. C. B. Dellaert, and K. D. Ruyter, International Journal of

Service Industry Management 15, 102 (2004).30. B. J. Corbitt, T. Thanasankit, and H. Yi, Electronic Commerce Research and

Applications 2, 203 (2003).31. Heijden, V. D. Hans, T. Verhagen, and M. Creemers, European Journal of

Information Systems 2, 41 (2003).32. www.utm.my.33. R. Krejcie and D. Morgan, Determining Sample Size for Research Activities:

Education and Physiological Measurement, Upper Saddle River, N. J, PrenticeHall (1970).

34. Y. Salaün and K. Flores, International Journal of Information Management21, 21 (2001).

35. D. Bisdee, Consumer Attitudes Review, Office of Fair Trading, June (2007),pp. 1–147.


3421




Vol. 21, 3422–3425, 2015

Characteristics of Trustees and Trustors Affecting

Consumer Trust in Online Purchasing

Nur Shafiqah Ghazali1, Mohd Shoki Md. Ariff1�∗, Khalid Ismail2, Abdul Halim Ali2,Amir Hasan Dawi2, and Nawawi Ishak3



In e-commerce, trust is a critical success factor influencing the acceptance of online purchasing. Online shop-ping involves both online sellers and buyers, and their characteristics are important in determining the success ofinternet-based businesses. Purchase intention behavior of consumers is greatly affected by trust and thereforeexamining characteristics of trustors and trustees influencing consumer trust is needed. In this study, mea-surements of consumer trust include both a trusting party (trustor) and a party to be trusted (trustee). Thecharacteristics of trustees are measured based on perceive reputation; perceive size, and system assuranceof online sellers, while characteristic of trustors (online consumers) is determined by propensity to trust. Self-administered questionnaire was employed to collect 250 (83% response rate) completed questionnaire frombusiness students of a public university in Malaysia. In online purchasing, perceived reputation and systemassurance of trustees and propensity to trust of trustors positively and significantly affected consumer trust. Per-ceives reputation exerted higher effect on consumer trust, highlighting how consumer perceived credibility andreputation of online sellers is vital in forming trust-related behavior in online shopping. Perceived size of onlinesellers business was found to have a negative and insignificant effect on consumer trust. Purchase intention ofconsumers was significantly and positively affected by their trust towards the online purchasing. Theoretical andpractical implications of the study were discussed.

Keywords: Consumer Trust, Commitment-Trust Theory, Theory of Planned Behavior, Purchase Intention.

1. INTRODUCTIONIn online purchasing, online buyers will be attracted to buy

products using electronic service channels if they trust the seller

or the website. Trust plays an important role in online purchasing,

and consumers’ purchase intention was positively influenced by

their trust towards the online sellers.1 According to Teo and Liu,2

online purchasing has limitation such as the physical separation

between consumer and online seller, and between consumer and

product or service. This limitation created many problems or

uncertainties in online purchasing, and one of the main prob-

lems is lack of consumer trust.3 Even though most people are

choosing to shop online, but some of them did not do it because

of they didn’t trust the online sellers. Online buyers who shop

online concern more on trust because they can’t touch and feel

the products, but they had to pay for the products before they

are sent to them. Some of online buyers did not trust online

shopping because they can’t touch and feel the product, issues

of security of the website and lack of trust on the characteris-

tics of online seller and their online shopping website.4 In order

to reduce this barrier, online sellers must develop a trustworthy


relationship to gain customer loyalty and long term relationship

with online buyers.

Previous researches addressing consumer trust in online setting

assessed this issue from technology perspective, such as per-

ceived ease of use and perceive usefulness,5–7 and perceive

credibility.7 Others view it from human behavior perspective, for

example perceived risks and trust in website.5�8 However, the

consumers’ trust in online purchasing also develops based on

how they judge characteristics of online sellers and their propen-

sity to trust.2 Chen and Dhillon9 suggest that consumers will

trust in online purchasing based on situation or belief that the

online vendor is reliable, i.e., how they perceive characteristics

of online sellers. The antecedent of consumer trust in the con-

text of online purchasing is based on characteristics of trustees

such as perceive reputation, perceive size, system assurance; and

characteristic of trustors which is propensity to trust.2 These char-

acteristics are related with the trust that effect customer decision

where the consumers’ attitudes toward online seller will affect

their willingness to buy. Both characteristics are important deter-

minant of trust in online purchasing. Therefore, examining char-

acteristics of trustees and trustors to determine consumer trust

should not be neglected.



This research is carried out to examine the characteristics of

trustees and trustors affecting consumer trust in online purchas-

ing and, to determine any intention to purchase if they trust the

online sellers. This study enriches literature on trust in online

shopping by examining both characteristics of online sellers and

online buyers to measure trust. The scope of the study is suitable

with both internet services technology and e-commerce disci-

plines with regard to consumer trust in internet technology-based

services. It goes beyond technology and human behavior factors

by focusing how online buyers judge characteristics of trustees

and trustors to form trust related behavior. For marketers, this

study contributes to the development of effective trust-building

marketing strategies in their businesses based on the characteris-

tics of trustees and trustors.

2. LITERATURE REVIEWOnline shopping, also known as internet shopping, electronic

shopping, online purchasing or internet buying, can be defined as

the process of purchasing goods and services over the internet.10

It involves examining, searching for, browsing for or looking at

a product to get more information with the possible intention of

purchasing on the internet.11

Trust can be defined as particularly relevance in conditions of

ignorance or uncertainty with respect to the unknown or unknow-

able actions.12 Chen and Dhillon9 state that trust is comparison

between intentions to accept based on expectation of intention

or behavior under risk. Trust is a feeling of mutual acceptance

between two parties where it develops from continuous physical

interaction and leads to long term acceptance and commitment.13

In online shopping, this involves the trustworthy relationship

between online sellers and online buyer, which requires trust of

consumers towards the sellers.

In the classic Theory of Commitment-Trust, the successful

relationship marketing requires relationship of commitment and

trust between seller and buyers.14 In online setting, trust is the

main factor why consumer purchases online and this is one of

the ways to maintain long term relationship between online sell-

ers and online buyers. The trust of consumers could be derived

from the characteristics of trustees or online sellers and the char-

acteristic of trustors or online consumers.2 Teo and Liu2 claim

that consumer trust in online purchasing is predicted by two

types of characteristics; first is characteristics of trustees or online

sellers, and second is characteristic of trustors or consumers.

The characteristics of trustees are perceived reputation and per-

ceived size2�15�16 and system assurance2�13�15–17 and multichannel

integration.2 Confident is the other dimension of trust, which is

arises from the online seller’s reputation. The confident in rep-

utation of online seller could be seen from the strength of the

brand name, endorsement from trusted third parties, and previous

interactions with consumers. System assurance is a technology

based-trust that influence the perceive reliability of the system.

Perceive size is based on the overall size and market share posi-

tion of online sellers,18 in which a large size shows that the sellers

are likely to possess expertise and necessary support system that

could encourage consumer trust.2

Based on the commitment-trust theory, the customer’s propen-

sity to trust the online seller,1�2�15 customer confident in the

website, and customer’s trust in internet technology features are

the key dimensions of trust in online purchasing. Propensity to

trust is important because the consumer and online seller are

physically separated, where the high level of satisfaction and

customer experience will have more trust in online purchasing.

Thus, propensity to trust is the major antecedent of trust in online

purchasing.1�2�13�16�17�19

Based on Theory of Planned Behavior (TPB),20 it was found

that trust on an online retailer was statistically significantly corre-

lated with attitudes toward online transactions and with perceived

behavioral control.21 Therefore, belief in the trustworthiness of

the online purchasing should be associated with a willingness to

purchase online or having purchase intention. Belief about the

trustworthiness toward online seller was associated with positive

attitudes toward online purchasing, and these positive attitudes

will in turn associated with having purchase intention thus lead

to actual purchasing behavior.11�21

Based on the commitment-trust theory and arguments on the

antecedents of trust, consumer trust is derived from the charac-

teristics of trustees or online sellers—perceived reputation (PR),

perceived size (PS) and System assurance (SA)–and the charac-

teristic of trustors or online consumers—Propensity to trust (PT).

In TPB, trustworthiness toward online seller was associated with

positive attitude toward online purchasing, and this positive atti-

tude will in turn associated with having purchase intention. Thus,

the framework of this study is proposed, as depicted in Figure 1.

Reputation is defined as the extent where the online consumers

believe or trust an online seller is a professionally competent or

honest, fair and benevolent.2 Reputation is important because it

is hard to form reputation than to lose it. Reputation could easily

damage if online seller did not carefully protect their reputation

in online business which it is a valuable asset that required a lot

of effort.2

A seller’s size (perceive size) is its overall size and market

share position.18 A large overall size and market share show that

the online seller consistently delivers on its promises to the con-

sumers and many consumers have believed it. Large size also

shows that the seller likely possess expertise and necessary sup-

port system that could encourage trust and consumer loyalty.2

Large size also suggests that online seller is able to assume the

risk of product failure or losses and compensate their consumer,

thus contribute to the consumer trust towards the online sellers.

System assurance can be defined as the dependability and

security of the online transaction system through the internet

which it is secure and successfully function.2 Due to uncer-

tainty in purchasing online, consumer will search on information

regarding the website and products or services offer by online

sellers. The information provided in a website must be consistent

and reliable11 and the website design must encourage consumers

to know more on the product or service offer. Online sellers

PurchaseIntention

ConsumerTrust

PR

PS

SA

PT

Characteristicsof trustors

Characteristicsof trustees

Fig. 1. The conceptual framework of the study.

3423


should know how to create perception of trust in consumers mind

when they are using the website. These features of system assur-

ance of a website positively affect trustworthy of consumer to

online vendor and therefore, the relationship between trustwor-

thiness and usage attitude can be enhanced.22

Propensity to trust is general willingness to trust others and it

is a measurement of an individual’s tendency to trust or distrust.2

Chen and Dhillon9 suggest that consumer will trust in online

purchasing based on situation or belief that the online vendor is

reliable. The propensity to trust in online purchasing will encour-

age consumer to develop trust and online purchase intention.

Several researches have shown a direct relationship between

trust and willingness to buy from online sellers. Kim et al.11

expect that increase in trust will directly and positively affect

purchase intention. It is important to measure the consumer trust

as it can be used to indicate the purchase intention among online

consumer.16

3. METHODOLOGYA questionnaire survey was employed to gather primary data on

characteristics of trustees and trustors, consumer trust and pur-

chase intention. In the questionnaire, 20 questions related to the

perceive reputation, perceive size, and system assurance of char-

acteristics of trustees were developed based on the work of Teo

and Liu;2 three question of consumer trust;2 and three questions

for purchase intention.11 Four questions regarding the characteris-

tics of trustors, that is propensity to trust, were constructed based

on the work of Teo and Liu.2 The scale of items was measured

based on five points Likert ranging from 1-strongly disagree to

5-strongly agree. All the questions must be answered based on

respondents’ experience of online purchasing with their selected

online vendors.

All students of a business faculty of a public university in

Malaysia serve as a population of the study. Students are a use-

ful surrogate for online consumers11 because they are generally

younger and more educated than conventional consumers. The

familiarity of university students, particularly business students

with the internet technology, e-commerce and online marketing is

suitable with their emerging market power as online consumers.

Further, it is easier to develop loyal consumer by establishing

initial relationship between online vendor and online buyers (the

students) in online purchasing.23 Using multivariate research,24

the sample size 300 is determined based on 5% margin error.

Convenience sampling was used because it is the most relevant

technique in sampling the target population where the total pop-

ulation is unknown (students who have online purchasing expe-

rience). 300 questionnaires were distributed to the students of

the Faculty with past online purchasing experience, and 250 sets

of the questionnaires were fully answered by the respondents.

Hence, the response rate of this study is 83.33%.

All variables used in this study are independence free and the

underlying assumptions of correlation and regression analyses

were met based on the normality test, linearity test and multi-

collinearity test performed. The KMO measure of sampling for

PR, PS, and SA of characteristics of trustees and PT of char-

acteristics of trustors, Consumer Trust, and Purchase Intention

are 0.731, 0.669 and 0.691 respectively, supported by Bartlett’s

test of Sphericity with significant values of 0.000 for all the

variables. These results of KMO and Barlett’s Test allow fac-

tor analysis to be performed. The Exploratory Factor Analysis

(EFA) performed yielded three components for characteristics of

trustees (PR, PS, and SA) and one for characteristics for trustors

(PT) with eigenvalues >1, with the total variance explained of

75.422%. For consumer trust and purchase intention, results of

EFA indicated that only one component for each variable with

eigenvalues >1, with the total variance explained of 63.609%

and 68.406% respectively. All questions for PR (except PR 1),

PS, SA, PT, consumer trust and purchase intention were retained

since the factor loadings of the variables were >0.5.

In order to measure the degree of consistency and correlations

among the items, reliability test was performed. The Cronbach’s

alpha values for the three characteristics of trustees were 0.826

(PR), 0.894 (PS) and 0.869 (SA); 0.832 (PT) of characteristics of

trustors; 0.711 (consumer trust) and 0.760 for purchase intention.

Thus, all the variables in the questionnaire were reliable and valid

for further analysis.

4. RESULT AND ANALYSISMultiple Regression analysis was performed to examine the effect

of perceive reputation (PR), perceive size (PS), system assurance

(SA), and propensity to trust (PT) on consumer trust (CT). As can

be seen in Table I, PR and SA of characteristics of trustees and

PT of trustors significantly affect consumer trust towards online

shopping. PR (� 0.384, t 6.772, Sig. 0.000) and SA (� 0.225,

t 4.149, Sig. 0.000) of characteristics of trustees positively affect

consumer trust. However, the effect of PS on consumer trust is

insignificant (�− 0�038� t − 0�708, Sig. 0.480). PT of character-

istic of trustors (� 0.356, t 7.150, Sig. 0.000) significantly and

positively affect consumer trust. The R2 value shows that 61.1%

of the variation of CT is explained by the variation of PR, PS, SA

and PT. Therefore, the linear regression equation for this study

was Yct = 0�400X1 +0�229X2 +0�352X3 +0�370.

Linear regression was performed to examine the effect of con-

sumer trust on purchase intention. Based on Table II, it shows

that purchase intention is significantly affected by consumer trust

(� 0.456, t 8.077, Sig. 0.000). Thus, consumer trust was found

to have positive and significant effect on purchase intention.

5. DISCUSSIONSConsumer will trust in online purchasing based on situation or

belief that the online vendor is reliable,9 and how reliable is

online sellers will be determined by characteristics of trustees and

trustors, as perceived by online buyers. The result of this study

shows that both characteristics are important determinants of trust

in online shopping. This finding is also observed in United State,

Singapore and China.2 Thus, in online shopping, characteristics

of trustees and trustors are equally important as technology and

Table I. Result on the effect of characteristics of trustees (PR, PS, and

SA) and characteristics of trustors (PT) on consumer trust.


Constant 0�370 0.180 2.058 0.041PR 0�400 0.059 0.384∗∗ 6.772 0.000 2.02PS −0�030 0.042 −0.038 −0.708 0.480 1.85SA 0�229 0.055 0.225∗∗ 4.149 0.000 1.86PT 0�352 0.049 0.356∗∗ 7.150 0.000 1.56F 96.368 R2 0.611

Notes: Dependent variable is consumer trust; ∗p < 0�05; ∗∗p < 0�01.

3424


Table II. Result on the effect of consumer trust on purchase intention.


Constant 1.749 0.247 7.094 0.000Consumer trust 0.511 0.063 0.456∗∗ 8.077 0.000 1.00F 65.233 R2 0.208

Notes: Dependent variable is purchase intention; ∗p < 0�05; ∗∗p < 0�01.

human behavior factors in understanding consumer trust. This

finding is consistent with many previous studies.2�9�11�22

Perceive reputation of online sellers significantly contribute to

consumer trust, which is consistent with previous studies.2�15�16

It shows that when online buyers perceived online sellers as

being honest, fair and consumer oriented, the trust related behav-

ior would be developed. From the system assurance perspective,

this research highlights that consumer still believe that security

of transaction system is important in online purchasing.2�13�15–17

Therefore, it is important for online sellers to ensure that the

online transaction systems must be stable, reliable, dependable

and secure. System approach deals with the extent to which

online buyers’ belief the online transaction of online sellers is

stable, reliable, dependable and secure,2�7 which is related to the

technology or websites quality of online sellers. Perceived rep-

utation relates to reputation of online sellers in term of having

good reputation and being honest, consumer oriented and faired

to online buyers, which is concerned to individual buyers. Inte-

grating these two perspectives provides a better understanding

of consumer trust in online setting. It implies that, in stimulat-

ing online trust, focusing on system assurance of technology or

website factor only is insufficient.

The propensity to trust of characteristic of trustors has positive

effect on consumer trust, which is consistent with the previ-

ous research findings.2�9 This study shows that, in online shop-

ping, individual’s general perceptions of whether or not they can

believe others play an important role in developing trust related

behavior. Thus, creating situation in which online buyers can eas-

ily develop trust towards the online sellers should be emphasized.

In addition, positive relationship exists between consumer

trusts with purchase intention. This finding synchronises with

what had been addressed by Kim et al.11 and Chen and Barnes.16

This shows that when consumers trust the online sellers, they will

have intention to make a purchase. The trust-purchase intention

relationship is observed in many studies worldwide, and therefore

it can be concluded that consumers’ trust precedes their inten-

tion to commit in online purchasing. E-marketers should develop

their marketing strategies based on the trust-intention relation-

ship, such as highlighting firms’ reputation in e-commerce setting

and build confident in the usage companies’ websites to stimulate

online purchasing.

6. CONCLUSIONSThis study enriches literature on trust in online shopping by

examining both characteristics of online sellers and online buy-

ers to determine consumer trust, as well as the effect of trust

on purchase intention. The finding suggests that addressing both

characteristics of trustees and trustors is important in under-

standing trust related behavior of consumer in online shopping.

For marketers, effective trust-building mechanisms based on the

characteristics of trustees and trustors should be embedded in

their business strategies.

This study addresses consumer trust from the business stu-

dents’ point of view, in which both characteristics of trustees

and trustors are important determinants of consumer trust in

online shopping. Business students are well exposed to both

e-commerce and internet-based information system; therefore,

their view on the role of trustees and trustors characteristics

in developing consumer trust should not be neglected. Future

research comparing the business students’ view with information

and communication technology students is interesting to study

because the later provide additional insight from information

technology literate consumers. Participation of business students

in this study is unlikely to present all the business students’ view;

therefore, similar study should be extended to cover as many as

possible business students in Malaysia higher education institu-

tions. Further, the negative and insignificant effect of perceive

size of trustees characteristics on consumer trust might be influ-

enced by the inability of respondents to imagine how big is the

online sellers’ market share. Therefore, it is important to objec-

tively show the market share of specific online sellers, so that,

respondents could make sound judgement on how perceive size

of the market share effect consumer trust.

References and Notes1. R. Connolly and F. Bannister, Management Research News 31, 339 (2008).2. T. S. H. Teo and J. Liu, Consumer trust in e-commerce in the United States,

Singapore and China, Omega (2007), Vol. 35, pp. 22–38.3. K.-H. Xiao, Journal of Advances in Information Sciences and Service

Sciences (AISS) 4, 170 (2012).4. Ecommerce Milo website, E-commerce infographic: Understanding

online shoppers in Malaysia. Available at: www.ecommercemilo.com/2014/01/ecommerce-infographic-malaysia-understanding-online-shoppers.html (2014).

5. H. van der Heijden, European Journal of Information Systems 12, 41.6. T. Li Wang and Y. F. Tseng, International Journal of Digital Society 2, 433

(2011).7. M. S. Md. Ariff, S. M. Yeow, and N. Zakuan, Adv. Sci. Lett. 20, 268 (2013).8. L. Zheng, M. Favier, P. Huang, and F. Coat, Journal of Electronic Commerce

Research 13, 255 (2012).9. S. C. Chen and G. S. Dhillon, Information Technology and Management 4, 303

(2003).10. Mastercard Worldwide Insights, Online shopping in Asia-Pacific-patterns,

trends and future growth, available at: www.mastercard.com/us/company/en/insights/studies/2008/asiaonlineshopping.htht (2008).

11. D. J. Kim and D. L. Ferrin, and H. R. Rao, Decision Support Systems 44, 544(2008).

12. D. G. Gambetta, Trust; Making and Breaking Corporative Relations, ElectronicEdition (1988), pp. 213–237.

13. S. Sahney, K. Ghosh, and A. Shrivastava, Journal of Asia Business Studies7, 278 (2013).

14. R. M. Morgan and S. D. Hunt, Journal of Marketing 58, 20 (1994).15. A. Mukherjee and P. Nath, European Journal of Marketing 41, 1173 (2007).16. Y.-H. Chen and S. Barnes, Industrial Management and Data Systems 107, 21

(2007).17. W. Gong, R. L. Stump, and L. M. Maddox, Journal of Asia Business Studies

7, 214 (2013).18. P. M. Doney and J. P. Cannon, Journal of Marketing 61, 35 (1997).19. C. Bianchi and L. Andrews, International Marketing Review 29, 253 (2012).20. I. Ajzen, Organizational Behavior and Human Decision Processes 50, 179

(1991).21. J. F. George, Internet Research 14, 198 (2004).22. P. Palvia, Information and Management 46, 213 (2009).23. Y. Xu and V. A. Paulins, Journal of Fashion Marketing and Management 9, 420

(2005).24. J. F. Hair, W. C. Black, B. J. Babin, R. E. Anderson, and R. L. Tatham,

Multivariate Data Analysis, 6th edn., Upper Saddle River, N. J., PearsonPrentice Hall (2006).


3425